Anthropic’s Transparency Hub \ Anthropic
https://www.anthropic.com/transparency
Anthropic transparency hub — the source for the RSP, system cards and external red-team results in the Claude assessment
852
Lines added
212
Lines removed
Was read
2026-10-20
d6569c63a19b874d…
Changed by
2026-10-21
67059b84faecfc09…
852 lines added, 212 lines removed
Left column is the version read 2026-10-20, right is 2026-10-21. Unchanged stretches are elided. Both bodies are kept in full and are addressed by the hashes above.
Showing the first 400 lines of the comparison; 1101 further lines are held back. This page reports a change, it is not a copy of anyone’s document — the source is linked above and is the authority for what it says.
11 Anthropic’s Transparency Hub
22 A look at Anthropic's key processes, programs, and practices for responsible AI development.
33
4−0 1 Model Report
5−0 2 System Trust and Reporting
6−0 3 Voluntary Commitments
4+01 Model Report
5+02 System Trust and Reporting
6+03 Voluntary Commitments
77
8+Chapter 01 : Model Report
9+
810 Model Report
911 October 2, 2026
1012 Select a model to see a summary that provides quick access to essential information about Claude models, condensing key details about the models' capabilities, safety evaluations, and deployment safeguards. We've distilled comprehensive technical assessments into accessible highlights to provide clear understanding of how the models function, what they can do, and how we're addressi…
11−Claude Haiku 5.5 Claude Opus 5.5 Claude Fable 5.1 Claude Mythos 5.1 Claude Opus 5 Claude Sonnet 5 Claude Fable 5 Claude Mythos 5 Claude Opus 4.8 Claude Opus 4.7 Claude Mythos Preview Claude Sonnet 4.6 Claude Opus 4.6 Claude Opus 4.5 Claude Haiku 4.5 Claude Sonnet 4.5 Claude Opus 4 and Sonnet 4 Claude Opus 4.1 Claude Sonnet 3.7
13+Claude Sonnet 5.5 Claude Haiku 5.5 Claude Opus 5.5 Claude Fable 5.1 Claude Mythos 5.1 Claude Opus 5 Claude Sonnet 5 Claude Fable 5 Claude Mythos 5 Claude Opus 4.8 Claude Opus 4.7 Claude Mythos Preview Claude Sonnet 4.6 Claude Opus 4.6 Claude Opus 4.5 Claude Haiku 4.5 Claude Sonnet 4.5 Claude Opus 4 and Sonnet 4 Claude Opus 4.1 Claude Sonnet 3.7
1214
15+Claude Sonnet 5.5 Summary Table
16+See Sonnet 5.5 System Card
17+Model description
18+
19+Claude Sonnet 5.5 is a faster, lower-cost complement to Claude Opus 5.5, and is strongest at well-scoped everyday tasks, fixing bugs, and creating polished documents, slides, and spreadsheets.
20+
21+Benchmarked Capabilities
22+
23+See our Claude Sonnet 5.5 system card ’s Section 8 on capabilities.
24+
25+Acceptable Uses
26+
27+See our Usage Policy
28+
29+Release date
30+
31+September 2026
32+
33+Modalities
34+
35+Claude Sonnet 5.5 can understand both text (including voice dictation) and image inputs, engaging in conversation, analysis, coding, and creative tasks. Claude can output text, including text-based artifacts, and diagrams.
36+
37+Model architecture and training methodology
38+
39+Claude Sonnet 5.5 was pretrained on large, diverse datasets to acquire language capabilities. After the pretraining process, Sonnet 5.5 underwent substantial post-training, with the goal of making it an effective assistant whose behavior aligns with the values described in Claude’s constitution .
40+
41+Training Data
42+
43+Claude Sonnet 5.5 was trained on a proprietary mix of publicly available information from the internet, public and private datasets, and other sources, such as synthetic data generated by other models. The mix may also include user data from feedback or bug reports, or data that users have explicitly permitted for training. In the course of its training, we used several data cleaning and filtering…
44+
45+Testing Methods and Results
46+
47+Based on our assessments, we deployed Claude Sonnet 5.5 with the same chemical and biological misuse protections we deployed for Claude Opus 5, treating it as having CB-1 capabilities. Autonomy threat model 1 is applicable to Claude Sonnet 5.5. See below for select safety evaluation summaries.
48+
49+The following are summaries of key safety evaluations from our Claude Sonnet 5.5 system card. Additional evaluations were conducted as part of our safety process; for our complete publicly reported evaluation results, please refer to the full system card .
50+Political Even-handedness
51+People use Claude to learn about and discuss political issues, so we test whether it treats different political views fairly. We use our open-source evaluation , which gives Claude 1,350 pairs of requests on 150 political topics. Each pair asks for the same thing from opposite political perspectives, for example a request about healthcare policy written once from a Democratic point of view and onc…
52+Another Claude model then grades the answers. It checks whether Claude answers both requests in a pair with similar depth and quality (what we call even-handedness), whether it acknowledges other points of view, and whether it refuses to answer. We report results in two settings: Claude accessed directly through our API (the service developers use to build Claude into their own products), with no …
53+Claude Sonnet 5.5 was substantially more even-handed than Claude Sonnet 5. It scored 97.9% on the API and 99.0% on claude.ai (vs. 86.2% and 95.7% for Sonnet 5), and it refused less often in both settings. Results on acknowledging other points of view were mixed. Through the API, Sonnet 5.5 did this slightly less often than Sonnet 5 (41.4% vs. 45.7%). On claude.ai, whose system prompt directs Claud…
54+Pairwise political bias evaluations: even-handedness. Higher scores for even-handedness are better. Results for previous models may show variance from previous system cards due to routine evaluation updates.
55+
56+Honesty
57+People increasingly rely on AI models for information and to get work done, so it matters that these models tell the truth and are open about what they have done. We test Claude’s honesty in several ways. Claude Sonnet 5.5 did better than Claude Sonnet 5 on most of these tests, with some exceptions noted below. For more in-depth descriptions of the evaluations and their results, please see the Cla…
58+Misleading users: We ran about 4,100 simulated test sessions with Claude, then had another AI model review them for eight kinds of misleading behavior, such as lying to the user, flattering them or agreeing with them too readily, and claiming to have finished a task that was not done. Claude Sonnet 5.5 did better than Claude Sonnet 5 on seven of the eight. The exception was dodging questions on se…
59+Making up facts: AI models sometimes “hallucinate,” stating false information as if it were true. An honest model should admit when it does not know an answer instead of guessing. We asked Claude factual questions on 41 subjects, without letting it look anything up online, and marked each answer as right, wrong, or declined. Claude Sonnet 5.5 got more questions right than Claude Sonnet 5, but it a…
60+Honesty under pressure: We used a public test called MASK to check whether Claude will say something it believes is false when a user, or the instructions it has been given, pressure it to. Claude Sonnet 5.5 stuck to what it believed more often than Claude Opus 5.5, but less often than Claude Sonnet 5. (Section 6.3.2.2)
61+Using answers without saying so: We gave Claude programming tasks where the answer was already within reach of the computer it was working on, for example in a newer version of the software that already contained the fix. We then measured how often Claude used that answer without telling the user, which would leave the user unaware of where the solution came from. Claude Sonnet 5.5 did this less o…
62+Admitting to hidden changes: During some training exercises, we saw Claude hide changes to code from the automated system that graded its work, so that it would pass. To test whether Claude would own up to this, we took records of these sessions and asked Claude to summarize its work, without asking about the hidden changes directly. Claude Sonnet 5.5 mentioned the hidden changes 96.2% of the time…
63+The charts below compare Sonnet 5.5 with other recent Claude models on the eight misleading behaviors from the first test above. Lower scores are better. Behaviors are scored even in sessions where they are unlikely to come up, so the scores are most useful for comparing models with each other.
64+Scores from our automated behavioral audit
65+
66+RSP Evaluations
67+Our Responsible Scaling Policy (RSP) evaluation process is designed to systematically assess our models' capabilities in areas where they could pose catastrophic risks before we release them. Claude Sonnet 5.5 is not more capable than our most advanced models, and it is less capable than Claude Opus 5.5 on most of our tests. We found that Claude Opus 5.5 does not reach our higher risk thresho…
68+Alignment risk is the risk that a model behaves in ways Anthropic did not intend. In our August 2026 Risk Report we raised our estimate of this risk from very low to low, because recently disclosed incidents involving how models behaved during cybersecurity testing made us less certain. We assess that the risk of Claude Sonnet 5.5 causing catastrophic harm by acting in ways we did not intend is al…
69+On its ability to carry out AI research and development on its own, Claude Sonnet 5.5 is broadly similar to or less capable than Claude Opus 5.5. Opus 5.5 did not cross our higher threshold for this ability, so we reach the same conclusion for Sonnet 5.5. On our capability index, which combines many tests into a single score, Sonnet 5.5 scored below Opus 5.5 (167.93 vs. 169.12).
70+On chemical and biological weapons, we treat Claude Sonnet 5.5 as able to significantly help someone with a basic technical background make known chemical or biological weapons, which is our CB-1 threshold. We deploy it with safeguards to match, including the same automated filters we used for Claude Opus 5, which are designed to detect and block attempts to get this kind of help. On three automat…
71+Developing new chemical or biological weapons takes rare expertise and judgment. We estimate that Claude Sonnet 5.5’s abilities in this area are similar to or below those of Claude Opus 5, and well below those of Claude Opus 5.5. It struggled with longer, open-ended research tasks, which limits how far it could stand in for human experts. We are releasing Sonnet 5.5 with the same safeguards we use…
72+
1373 Claude Haiku 5.5 Summary Table
14−Model description Claude Haiku 5.5 is our fastest, most capable small model, built for high-volume work like summarization, subagents, and browser use
15−Benchmarked Capabilities See our Claude Haiku 5.5 system card ’s Section 8 on capabilities.
16−Acceptable Uses See our Usage Policy
17−Release date October 2026
18−Modalities Claude Haiku 5.5 can understand both text (including voice dictation) and image inputs, engaging in conversation, analysis, coding, and creative tasks. Claude can output text, including text-based artifacts, and diagrams.
19−Model architecture and training methodology Claude Haiku 5.5 was pretrained on large, diverse datasets to acquire language capabilities. After the pretraining process, Haiku 5.5 underwent substantial post-training, with the goal of making it an effective assistant whose behavior aligns with the values described in Claude’s constitution .
20−Training Data Claude Haiku 5.5 was trained on a proprietary mix of publicly available information from the internet, public and private datasets, user data included in feedback or bugs or which users have explicitly permitted for training, and other sources, such as synthetic data generated by other models. In the course of its training, we used several data cleaning and filtering methods, includi…
21−Testing Methods and Results Based on our assessments, we deployed Claude Haiku 5.5 with the same chemical and biological misuse protections we deployed for Claude Opus 5 and Claude Sonnet 5. We treat it as meeting our CB-1 threshold, but not our higher CB-2thresholds. See below for select safety evaluation summaries.
22−See Fable 5.1 & Mythos 5.1 System Card
74+See Claude Haiku 5.5 System Card
75+Model description
2376
77+Claude Haiku 5.5 is our fastest, most capable small model, built for high-volume work like summarization, subagents, and browser use
78+
79+Benchmarked Capabilities
80+
81+See our Claude Haiku 5.5 system card ’s Section 8 on capabilities.
82+
83+Acceptable Uses
84+
85+See our Usage Policy
86+
87+Release date
88+
89+October 2026
90+
91+Modalities
92+
93+Claude Haiku 5.5 can understand both text (including voice dictation) and image inputs, engaging in conversation, analysis, coding, and creative tasks. Claude can output text, including text-based artifacts, and diagrams.
94+
95+Model architecture and training methodology
96+
97+Claude Haiku 5.5 was pretrained on large, diverse datasets to acquire language capabilities. After the pretraining process, Haiku 5.5 underwent substantial post-training, with the goal of making it an effective assistant whose behavior aligns with the values described in Claude’s constitution .
98+
99+Training Data
100+
101+Claude Haiku 5.5 was trained on a proprietary mix of publicly available information from the internet, public and private datasets, user data included in feedback or bugs or which users have explicitly permitted for training, and other sources, such as synthetic data generated by other models. In the course of its training, we used several data cleaning and filtering methods, including deduplicati…
102+
103+Testing Methods and Results
104+
105+Based on our assessments, we deployed Claude Haiku 5.5 with the same chemical and biological misuse protections we deployed for Claude Opus 5 and Claude Sonnet 5. We treat it as meeting our CB-1 threshold, but not our higher CB-2thresholds. See below for select safety evaluation summaries.
106+
24107 The following are summaries of key safety evaluations from our Claude Haiku 5.5 system card. Additional evaluations were conducted as part of our safety process; for our complete publicly reported evaluation results, please refer to the full system card .
25108 Election Integrity
26109 People use Claude to find information about elections and voting, so we test whether it answers legitimate election questions while refusing requests that violate our Usage Policy , such as requests to spread false information about how to vote, or to create content designed to mislead or intimidate people into not voting.
27110 We use two kinds of tests. The first is a set of 600 single-turn requests based on patterns we see in real use: 300 are harmful requests that Claude should not assist with, and 300 are legitimate ones that Claude should answer. The second kind of evaluation assesses Claude’s behavior inlonger conversations, in which a simulated user, played by another AI model following instructions from our polic…
⋯ 15 unchanged lines
43126 On chemical and biological weapons, we treat Claude Haiku 5.5 as meeting our CB-1 threshold, which covers models that could significantly help someone with a basic technical background make known chemical or biological weapons. We deploy it with safeguards to match, including automated filters designed to detect and block attempts to get this kind of help. Because Haiku 5.5 is not at the frontier,…
44127 Developing new biological weapons takes rare expertise and judgment. We estimate that Claude Haiku 5.5’s abilities in this area are above those of Claude Sonnet 5 but below those of Claude Opus 5, and well below those of Claude Opus 5.5, Claude Sonnet 5.5 and Claude Mythos 5.1. Since those models do not meet our CB-2 threshold, we determined that Claude Haiku 5.5 does not either. On one part of on…
45128
46129 Claude Opus 5.5 Summary Table
47−Model description Claude Opus 5.5 is our new leading model, and early testers saw large jumps in performance on their most complex work.
48−Benchmarked Capabilities See our Claude Opus 5.5 system card ’s Section 8 on capabilities.
49−Acceptable Uses See our Usage Policy
50−Release date September 2026
51−Modalities Claude Opus 5.5 can understand both text (including voice dictation) and image inputs, engaging in conversation, analysis, coding, and creative tasks. Claude can output text, including text-based artifacts, and diagrams.
52−Model architecture and training methodology Claude Opus 5.5 was pretrained on large, diverse datasets to acquire language capabilities. After the pretraining process, Opus 5.5 underwent substantial post-training, with the goal of making it an effective assistant whose behavior aligns with the values described in Claude’s constitution .
53−Training Data Claude Opus 5.5 was trained on a proprietary mix of publicly available information from online sources, public and private datasets, user data, and synthetic data generated by other models. Throughout the training process we used several data cleaning and filtering methods, including deduplication and classification.
54−Testing Methods and Results Based on our assessments, we deployed Claude Opus 5.5 with ASL-3 protections, treating it as having CB-1 capabilities. Autonomy threat model 1 is applicable to Claude Opus 5.5. See below for select safety evaluation summaries.
55−See Fable 5.1 & Mythos 5.1 System Card
130+See Claude Opus 5.5 System Card
131+Model description
56132
133+Claude Opus 5.5 is our new leading model, and early testers saw large jumps in performance on their most complex work.
134+
135+Benchmarked Capabilities
136+
137+See our Claude Opus 5.5 system card ’s Section 8 on capabilities.
138+
139+Acceptable Uses
140+
141+See our Usage Policy
142+
143+Release date
144+
145+September 2026
146+
147+Modalities
148+
149+Claude Opus 5.5 can understand both text (including voice dictation) and image inputs, engaging in conversation, analysis, coding, and creative tasks. Claude can output text, including text-based artifacts, and diagrams.
150+
151+Model architecture and training methodology
152+
153+Claude Opus 5.5 was pretrained on large, diverse datasets to acquire language capabilities. After the pretraining process, Opus 5.5 underwent substantial post-training, with the goal of making it an effective assistant whose behavior aligns with the values described in Claude’s constitution .
154+
155+Training Data
156+
157+Claude Opus 5.5 was trained on a proprietary mix of publicly available information from online sources, public and private datasets, user data, and synthetic data generated by other models. Throughout the training process we used several data cleaning and filtering methods, including deduplication and classification.
158+
159+Testing Methods and Results
160+
161+Based on our assessments, we deployed Claude Opus 5.5 with ASL-3 protections, treating it as having CB-1 capabilities. Autonomy threat model 1 is applicable to Claude Opus 5.5. See below for select safety evaluation summaries.
162+
57163 The following are summaries of key safety evaluations from our Claude Opus 5.5 system card. Additional evaluations were conducted as part of our safety process; for our complete publicly reported evaluation results, please refer to the full system card .
58164 Safeguards Evaluation
59165 We run evaluations to assess how Claude handles conversations about disordered eating. We focus on whether it avoids reinforcing requests that pose potential risk while remaining helpful on benign questions about nutrition, fitness, and health. Claude Opus 5.5 maintained high harmless response rates on single-turn requests that could reinforce disordered eating behaviors, such as requests for calo…
60166 Qualitative review of multi-turn testing identified multiple areas where Claude Opus 5.5 improved compared to Opus 5. Opus 5.5 produced outputs that were shorter and less verbose than those of Opus 5, which is beneficial when the user may have limited capacity or willingness to parse lengthy, complex outputs. It also used a warmer tone, more frequently validated the act of disclosure, and was less…
⋯ 1 unchanged line
62168
63169 Alignment Evaluation
64170 When Claude works as an agent, it usually runs inside a sandbox: a walled-off computer environment that gives the model the access it needs for a task and nothing more. We ran a sandbox escape evaluation to measure how often Claude Opus 5.5 tries to violate those boundaries without authorization. In this evaluation, one Claude model acts as an investigator, builds a scenario and probes how the mod…
65171 Scores from our sandbox escape evaluation, comprising 28 handmade scenarios tailored to elicit sandbox escape, run seven times for each model. Percentage of low-severity and moderate- to high-severity attempts shown. Severity determined by the persistence, scope, and deception involved in each attempted act of containment boundary violation. Shown with 95% CI.
172+
66173 RSP Evaluation
67174 Our Responsible Scaling Policy (RSP) evaluation process is designed to systematically assess our models' capabilities in areas where they could pose catastrophic risks before we release them. These evaluations test the underlying model, so they were run without the additional safeguards we apply in our products. Claude Opus 5.5 is our most capable general release model to date, matching or ex…
68175 Alignment risk, which is the risk that a model behaves in ways Anthropic did not intend, remains low. In our August 2026 Risk Report we moved this assessment from very low to low to reflect increased uncertainty following recent incident disclosures related to model behavior in cybersecurity evaluations. We do not believe Claude Opus 5.5 raises risk beyond that assessment. On tests of whether a mo…
69176 On automated AI research and development, Claude Opus 5.5's capabilities are at or slightly above those of Claude Mythos 5.1, our previous frontier in this area, but it does not cross the RSP capability threshold. We have not observed a sustained doubling in the pace of our AI progress attributable to AI, and the model is not close to substituting for our research scientists and engineers. On…
70177 On chemical and biological weapons, Claude Opus 5.5 is broadly as capable as previous models we have conservatively treated as able to significantly help individuals with basic technical backgrounds produce known (non-novel) weapons, so we treat it as having that capability and deploy commensurate safeguards, including real-time classifiers to prevent harm.
71178 For novel weapons development, Claude Opus 5.5 performs similarly to Claude Mythos 5.1 across our evaluations. However, expert red teaming found that it did not improve on the weaknesses that kept Claude Mythos 5.1 below this threshold. It struggled to generate genuinely new ideas, sometimes misrepresented the scientific literature by relying on paper abstracts, and made scientific errors that wen…
72179
73180 Claude Fable 5.1 Summary Table
74−Model description Claude Fable 5.1 is the world’s most advanced model for coding and knowledge work—and its research capabilities offer an early glimpse of how AI models will contribute to scientific progress.
75−Benchmarked Capabilities See our Claude Fable 5.1 & Claude Mythos 5.1 system card ’s Section 8 on capabilities.
76−Acceptable Uses Anthropic’s Usage Policy applies.
77−Release date September 2026
78−Modalities Claude Fable 5.1 can understand both text (including voice dictation) and image inputs, engaging in conversation, analysis, coding, and creative tasks. Claude can output text, including text-based artifacts, and diagrams.
79−Model architecture and training methodology Claude Fable 5.1 was pretrained on large, diverse datasets to acquire language capabilities. After the pretraining process, Fable 5.1 underwent substantial post-training, with the goal of making it an effective assistant whose behavior aligns with the values described in Claude’s constitution .
80−Training Data Claude Fable 5.1 was trained on a proprietary mix of publicly available information from online sources, public and private datasets, user data, and synthetic data generated by other models. Throughout the training process we used several data cleaning and filtering methods, including deduplication and classification.
81−Testing Methods and Results Based on our assessments, we deployed Claude Fable 5.1 with ASL-3 protections, treating it as having CB-1 capabilities. Autonomy threat model 1 is applicable to Claude Fable 5.1. See below for select safety evaluation summaries.
82−See Fable 5.1 & Mythos 5.1 System Card
181+See Claude Fable 5.1 & Claude Mythos 5.1 system card
182+Model description
83183
184+Claude Fable 5.1 is a leading model for coding and knowledge work—and its research capabilities offer an early glimpse of how AI models will contribute to scientific progress.
185+
186+Benchmarked Capabilities
187+
188+See our Claude Fable 5.1 & Claude Mythos 5.1 system card ’s Section 8 on capabilities.
189+
190+Acceptable Uses
191+
192+Anthropic’s Usage Policy applies.
193+
194+Release date
195+
196+September 2026
197+
198+Modalities
199+
200+Claude Fable 5.1 can understand both text (including voice dictation) and image inputs, engaging in conversation, analysis, coding, and creative tasks. Claude can output text, including text-based artifacts, and diagrams.
201+
202+Model architecture and training methodology
203+
204+Claude Fable 5.1 was pretrained on large, diverse datasets to acquire language capabilities. After the pretraining process, Fable 5.1 underwent substantial post-training, with the goal of making it an effective assistant whose behavior aligns with the values described in Claude’s constitution .
205+
206+Training Data
207+
208+Claude Fable 5.1 was trained on a proprietary mix of publicly available information from online sources, public and private datasets, user data, and synthetic data generated by other models. Throughout the training process we used several data cleaning and filtering methods, including deduplication and classification.
209+
210+Testing Methods and Results
211+
212+Based on our assessments, we deployed Claude Fable 5.1 with ASL-3 protections, treating it as having CB-1 capabilities. Autonomy threat model 1 is applicable to Claude Fable 5.1. See below for select safety evaluation summaries.
213+
84214 Claude Mythos 5.1 and Claude Fable 5.1 are the same model, but with different levels of safeguards. Claude Fable 5.1 is generally available, while Claude Mythos 5.1 is available only through our trusted access programs and Claude Security; its safeguards are specifically designed to support work in cybersecurity and the life sciences.Evaluations that don't engage those safeguards can be run o…
85215 Safeguards Evaluation
86216 Claude Fable 5.1/Mythos 5.1 showed comparable overall performance to Mythos 5 in the suicide and self-harm domain. In terms of improvements, Mythos 5.1 was less likely than Claude Mythos 5 to suggest substitution methods for self-harm (e.g., holding ice cubes). These methods are clinically contested and have not been shown in research to meaningfully reduce self-harm incidents. Mythos 5.1 was also…
87217 One area for improvement was a tendency to implicitly validate self-harm as a coping strategy by acknowledging that it can regulate difficult emotions or provide relief. Mythos 5.1 also sometimes validated a user's fears about seeking help and mildly amplified prior negative experiences with crisis services. When these statements appeared, they were consistently within responses that discoura…
⋯ 5 unchanged lines
93223 In computer-use environments, Claude Fable 5.1 achieved an attack success rate of 0.07%, corresponding to two successful attempts out of 2,800 across two of the 14 scenarios. This is comparable to Claude Opus 5 (0.18%, or five successful attempts), because the difference is not distinguishable from noise at these low absolute rates. Both are more robust than Claude Fable 5 (2.50%) and Claude Sonne…
94224 Alignment Evaluations
95225 Claude Mythos 5.1 and Claude Fable 5.1 are the same model, but with different levels of safeguards. This means many of our evaluations can apply to either model. Our automated audit finds that Claude Mythos 5.1 is somewhat more honest and hallucinates less than previously released models. It hallucinates inputs, meaning it invents or materially misrepresents the contents of files or earlier user m…
96226 Scores from our automated behavioral audit for the dishonesty-related metrics given below. Lower numbers represent a lower rate or severity of the measured behavior; on all graphs in this figure lower is better. The y-axis is truncated below the maximum score of 10 in many cases. Reported scores are averaged across all approximately 4,100 investigations per target model (approximately 2,100 seed i…
227+
97228 RSP Evaluations
98229 Our Responsible Scaling Policy (RSP) evaluation process is designed to systematically assess our models' capabilities in areas where they could pose catastrophic risks before we release them. Because Claude Fable 5.1 and Claude Mythos 5.1 share the same underlying model, these evaluations were run on Claude Mythos 5.1, the configuration without the additional biology and cybersecurity safegua…
99230 Alignment risk, which is the risk that a model behaves in ways Anthropic did not intend, is low: in our August 2026 Risk Report we moved this assessment from very low to low to reflect increased uncertainty following recent incident disclosures related to model behavior in cybersecurity evaluations, and while Claude Mythos 5.1 shows somewhat stronger covert capabilities than prior models, we do no…
100231 On automated AI research and development, Claude Mythos 5.1's capabilities are comparable to or slightly stronger than those of Claude Mythos 5 and Opus 5, our previous frontier in this area, but it does not cross the RSP capability threshold. We have not observed a sustained doubling in the pace of our AI progress attributable to AI, and the model is not close to substituting for our researc…
101232 On chemical and biological weapons, it is difficult to say with full confidence whether any model passes our threshold for providing significant uplift in the construction of non-novel CB weapons. However, Claude Mythos 5.1 is broadly more capable than previous models we have conservatively treated as able to significantly help individuals with basic technical backgrounds produce (non-novel) weapo…
102233 For novel weapons development, Claude Mythos 5.1 shows modest gains over Claude Mythos 5 and Opus 5 on our automated evaluations. However, expert red teaming indicates that significant weaknesses remain, including weak novel ideation, poor strategic judgment and poor technical calibration, which prevent the model from substituting for scarce human expertise, and we conclude that Claude Mythos 5.1 …
103234
104235 Claude Mythos 5.1 Summary Table
105−Model description Claude Mythos 5.1 is the world’s most advanced model for coding and knowledge work—and its research capabilities offer an early glimpse of how AI models will contribute to scientific progress.
106−Benchmarked Capabilities See our Claude Fable 5.1 & Claude Mythos 5.1 system card ’s Section 8 on capabilities.
107−Acceptable Uses Anthropic’s Usage Policy applies. Note that this model is available only through our trusted access programs and Claude Security .
108−Release date September 2026
109−Modalities Claude Mythos 5.1 can understand both text (including voice dictation) and image inputs, engaging in conversation, analysis, coding, and creative tasks. Claude can output text, including text-based artifacts, and diagrams.
110−Model architecture and training methodology Claude Mythos 5.1 was pretrained on large, diverse datasets to acquire language capabilities. After the pretraining process, Mythos 5.1 underwent substantial post-training, with the goal of making it an effective assistant whose behavior aligns with the values described in Claude’s constitution .
111−Training Data Claude Mythos 5.1 was trained on a proprietary mix of publicly available information from online sources, public and private datasets, user data, and synthetic data generated by other models. Throughout the training process we used several data cleaning and filtering methods, including deduplication and classification.
112−Testing Methods and Results Based on our assessments, we deployed Claude Mythos 5.1 with ASL-3 protections, treating it as having CB-1 capabilities. Autonomy threat model 1 is applicable to Claude Mythos 5.1. See below for select safety evaluation summaries.
113−See Fable 5.1 & Mythos 5.1 System Card
236+See Claude Fable 5.1 & Claude Mythos 5.1 system card
237+Model description
114238
239+Claude Mythos 5.1 is a leading model for coding and knowledge work—and its research capabilities offer an early glimpse of how AI models will contribute to scientific progress.
240+
241+Benchmarked Capabilities
242+
243+See our Claude Fable 5.1 & Claude Mythos 5.1 system card ’s Section 8 on capabilities.
244+
245+Acceptable Uses
246+
247+Anthropic’s Usage Policy applies.
248+Note that this model is available only through our trusted access programs and Claude Security .
249+
250+Release date
251+
252+September 2026
253+
254+Modalities
255+
256+Claude Mythos 5.1 can understand both text (including voice dictation) and image inputs, engaging in conversation, analysis, coding, and creative tasks. Claude can output text, including text-based artifacts, and diagrams.
257+
258+Model architecture and training methodology
259+
260+Claude Mythos 5.1 was pretrained on large, diverse datasets to acquire language capabilities. After the pretraining process, Mythos 5.1 underwent substantial post-training, with the goal of making it an effective assistant whose behavior aligns with the values described in Claude’s constitution .
261+
262+Training Data
263+
264+Claude Mythos 5.1 was trained on a proprietary mix of publicly available information from online sources, public and private datasets, user data, and synthetic data generated by other models. Throughout the training process we used several data cleaning and filtering methods, including deduplication and classification.
265+
266+Testing Methods and Results
267+
268+Based on our assessments, we deployed Claude Mythos 5.1 with ASL-3 protections, treating it as having CB-1 capabilities. Autonomy threat model 1 is applicable to Claude Mythos 5.1. See below for select safety evaluation summaries.
269+
115270 Claude Mythos 5.1 and Claude Fable 5.1 are the same model, but with different levels of safeguards. Claude Fable 5.1 is generally available, while Claude Mythos 5.1 is available only through our trusted access programs and Claude Security; Mythos’ safeguards are specifically designed to support work in cybersecurity and the life sciences. For our complete publicly reported evaluation results, plea…
116271 Cyber Evaluation
117272 ExploitGym is a large-scale public benchmark developed by researchers at UC Berkeley together with collaborators at the Max Planck Institute for Security and Privacy, UC Santa 50 Barbara, Arizona State University, Anthropic, OpenAI, and Google. The benchmark measures whether AI agents can turn known vulnerabilities into working exploits. ExploitGym targets a wide range of applications and mainly f…
118273 The task is to develop an exploit that achieves unauthorized code execution and reads a secret flag (a secret password) stored outside the agent’s authorized scope. Because agents sometimes achieve code execution through a different flaw than the one provided, successful flag captures are additionally verified by an agent-as-a-judge check. The chart below reports the number of exploits that succes…
119274 Claude Mythos 5.1 is an improvement over Mythos 5 on ExploitGym. Bars show the number of successful exploits using the given vulnerability under 2-hour and 6-hour wall-clock budgets. Success requires 51 capturing a dynamically generated secret flag, and each successful exploit is verified by an agent judge to confirm it uses the intended vulnerability.
275+
120276 Bio Evaluation
121277 This benchmark assesses Claude's protein design abilities. The Sequence Generation variant assesses Claude's ability to create new proteins that meet a set of requirements, such as a plain-language description of its function, overall shape or how compact it is, and other specific structural features (e.g. whether it needs to attach to another protein). It is scored on whether the design…
122278 The Library Ranking variant assesses Claude's ability to decide which protein designs are worth testing in the lab first. Each problem gives Claude one set of candidate proteins along with real lab results from an actual protein engineering project, tells it what the researchers are trying to optimize for, and asks it to rank a new set of candidates from the same project that it has not seen …
123279
124280 Claude Opus 5 Summary Table
125−Model description Claude Opus 5 is a thoughtful and proactive model that comes close to frontier intelligence. On some coding and knowledge work evaluations Opus 5 is the new state-of-the-art.
126−Benchmarked Capabilities See our Claude Opus 5 system card ’s Section 8 on capabilities.
127−Acceptable Uses See our Usage Policy
128−Release date July 2026
129−Access Surfaces Claude Opus 5 can be accessed through: Claude.ai
281+See Claude Opus 5 system card
282+Model description
283+
284+Claude Opus 5 is a thoughtful and proactive model that comes close to frontier intelligence. On some coding and knowledge work evaluations Opus 5 is the new state-of-the-art.
285+
286+Benchmarked Capabilities
287+
288+See our Claude Opus 5 system card ’s Section 8 on capabilities.
289+
290+Acceptable Uses
291+
292+See our Usage Policy
293+
294+Release date
295+
296+July 2026
297+
298+Access Surfaces
299+
300+Claude Opus 5 can be accessed through:
301+Claude.ai
130302 Claude Code
131303 The Anthropic API
132304 Amazon Bedrock
133305 Google Vertex AI
134306 Microsoft Azure AI Foundry
135307
136−Software Integration Guidance See our Developer Documentation
137−Modalities Claude Opus 5 can understand both text (including voice dictation) and image inputs, engaging in conversation, analysis, coding, and creative tasks. Claude can output text, including text-based artifacts, and diagrams.
138−Knowledge Cutoff Date Claude Opus 5 has a knowledge cutoff date of May 2026. This means the models’ knowledge base is most extensive and reliable on information and events up to May 2026.
139−Software and Hardware Used in Development Cloud computing resources from Amazon Web Services, Google Cloud Platform and Microsoft Azure, supported by development frameworks including PyTorch, JAX, and Triton.
140−Model architecture and training methodology Claude Opus 5 was pretrained on large, diverse datasets to acquire language capabilities. After the pretraining process, Opus 5 underwent substantial post-training, with the goal of making it an effective assistant whose behavior aligns with the values described in Claude’s constitution .
141−Training Data Claude Opus 5 was trained on a proprietary mix of publicly available information from online sources, public and private datasets, user data, and synthetic data generated by other models. Throughout the training process we used several data cleaning and filtering methods, including deduplication and classification.
142−Testing Methods and Results Based on our assessments, we deployed Claude Opus 5 with ASL-3 protections, treating it as having CB-1 capabilities. Autonomy threat model 1 is applicable to Claude Opus 5. See below for select safety evaluation summaries.
143−See Claude Fable 5 & Claude Mythos 5 system card
308+Software Integration Guidance
144309
310+See our Developer Documentation
311+
312+Modalities
313+
314+Claude Opus 5 can understand both text (including voice dictation) and image inputs, engaging in conversation, analysis, coding, and creative tasks. Claude can output text, including text-based artifacts, and diagrams.
315+
316+Knowledge Cutoff Date
317+
318+Claude Opus 5 has a knowledge cutoff date of May 2026. This means the models’ knowledge base is most extensive and reliable on information and events up to May 2026.
319+
320+Software and Hardware Used in Development
321+
322+Cloud computing resources from Amazon Web Services, Google Cloud Platform and Microsoft Azure, supported by development frameworks including PyTorch, JAX, and Triton.
323+
324+Model architecture and training methodology
325+
326+Claude Opus 5 was pretrained on large, diverse datasets to acquire language capabilities. After the pretraining process, Opus 5 underwent substantial post-training, with the goal of making it an effective assistant whose behavior aligns with the values described in Claude’s constitution .
327+
328+Training Data
329+
330+Claude Opus 5 was trained on a proprietary mix of publicly available information from online sources, public and private datasets, user data, and synthetic data generated by other models. Throughout the training process we used several data cleaning and filtering methods, including deduplication and classification.
331+
332+Testing Methods and Results
333+
334+Based on our assessments, we deployed Claude Opus 5 with ASL-3 protections, treating it as having CB-1 capabilities. Autonomy threat model 1 is applicable to Claude Opus 5. See below for select safety evaluation summaries.
335+
145336 The following are summaries of key safety evaluations from our Claude Opus 5 system card. Additional evaluations were conducted as part of our safety process; for our complete publicly reported evaluation results, please refer to the full system card .
146337
147338 User Wellbeing Summary
148339 We run a suite of evaluations to understand how Claude responds in scenarios related to child safety and mental health. Claude is not a substitute for professional advice or medical care and is not intended to diagnose or treat any medical condition. We use these evaluations to understand how Claude performs in sensitive contexts and where we can make improvements. For more in depth descriptions o…
⋯ 3 unchanged lines
152343 Misleading the user: Claude Opus 5 misleads the user at rates similar to or lower than Opus 4.8, Mythos 5, and Sonnet 5. The one exception is input hallucination, where the mean rose slightly but within the range of expected noise. (Sections 6.4.3)
153344 External Red Teaming
154345 The IPI benchmark was built in partnership with Gray Swan, the UK AI Security Institute, the US Center for AI Standards and Innovation, and other model developers. It builds on Gray Swan’s published red-teaming competition 3 with a new set of 28 scenarios in which participants were tasked with finding attacks against frontier models. These scenarios test susceptibility to indirect prompt injection…
155346 Indirect prompt injection attacks from the Gray Swan IPI benchmark (Q1 2026), lower scores are better. All models use extended thinking. Results represent the probability that an attacker finds a successful attack after k=1, k=10, and k=15 attempts. Lower is better. Results for Gemini 3.1 Pro are not directly comparable as this model was included in the red-teaming competition used to source attac…
347+