ANTHROPIC · MODEL RELEASE TRACKER
Claude Fable 5.1
Claude Fable 5.1 is described in supplied reporting as Anthropic’s more advanced Claude model, noted for strong safety safeguards (including classifiers for biology, cybersecurity, and AI-development domains) and as a high-performance reference point in benchmark comparisons.CURRENT SNAPSHOT5/5 DIMENSIONS WITH DATA
The dimensions that change the decision.
RELEASE TIMELINE
Published, source-backed release events only.
Anthropic releases Claude Fable 5.2; Claude Mythos 5.2 available to vetted orgs
Anthropic announced two new models, Claude Fable 5.2 and Claude Mythos 5.2, claiming improvements in coding, reasoning, and vision and faster output than the 5.1 versions. Anthropic says Fable 5.2 is broadly available today while Mythos 5.2 access is limited to vetted organizations, and that prerelease testing helped solve complex engineering and scientific tasks (including work on black hole physics and formalizing Fermat’s Last Theorem).
Google DeepMind unveils Gemini 4 Argon, a frontier model with 1M-token context
Google DeepMind announced Gemini 4 Argon, a new frontier multimodal model optimized for long-horizon reasoning and complex workflows. Argon is rolling out initially to trusted cyber defenders via the Fairwind Program, expands context length to 1 million tokens, reports leading benchmark performance across coding, finance, legal, and video understanding, and will be made more widely available after phased safety testing and engagement with U.S. government pre-release processes; Google also published introductory pricing.
SpaceXAI releases Grok 4.7 — faster, cheaper model for coding and knowledge work
SpaceXAI released Grok 4.7, a new base-model upgrade positioned as its most capable model for coding and knowledge work; it reportedly uses a larger base model, longer RL training focused on long-running tasks, and an entirely new safeguard stack. The company says Grok 4.7 matches the price and speed of Grok 4.6 while improving verification, longer-context handling, document/presentation generation, benchmark performance (CursorBench 4.0, GDPval, AA Briefcase), and safety metrics (LatchBio biosafety 62.4%, HackerBench v0.3 allowing 3.3% risky prompts); it is available in Cursor, Grok Build, the Grok API, third-party harnesses and cloud platforms, priced from $2 per million input tokens and $6 per million output tokens, with an optional faster (2x) variant at double the price.
VERIFIABLE FACTS
Every value stays attached to a source and date.
2026-09-01
1,000,000 tokens
128K tokens
Text, images
Text
Active on Claude API, Amazon Bedrock, Google Cloud, Microsoft Foundry and Claude Platform on AWS. Active (latest)
$10 USD / 1M input tokens; Claude API standard, global routing
$50 USD / 1M output tokens; Claude API standard, global routing
$0.25 USD / 1M cached input tokens; Claude API standard, global routing
Anthropic states Claude Fable 5.1 has conservative cyber safeguards that block most cyber work to limit harmful activities.
Described as having safeguards that re-route certain cybersecurity-related requests to less-powerful models and route biology-related flagged requests to other models.
Reportedly used as a performance reference: Anthropic said Opus 5.5 'matches Claude Fable 5.1 on most tasks' while costing about 40% less.
Some reporting states that Opus 5.5 outpaced the larger Fable model in many benchmarks.
Reported score of 70% on the Furniture Assembly Benchmark (FAB), placing Claude Fable 5.1 behind GPT-6 Astra (80%) and ahead of Claude Opus 5 (61%), according to The Decoder.
Claude Fable 5.1 was one of six models tested by Vals AI (in an experiment by Matilda Glynn-Henley) that simulated AI agents trying to secure a £350 train refund.
In the experiment Claude Fable 5.1 was among the models that had web and email access, allowing agents to search the web and send emails autonomously.
Each agent in the experiment, including Claude Fable 5.1, was allowed up to 40 million tokens.
Agents in the experiment, including Claude Fable 5.1, were allowed up to twelve simulated weeks to pursue the refund.
Solved 60% of tasks on CloudOpsBench v1.0 (45 AWS DevOps tasks).
WHAT CHANGED
Stored passport versions, without reconstructed history.
No public fact values changed.
No public fact values changed.