Every figure, price, date and benchmark result the video puts on screen, chased to a primary source.
Anthropic released Claude Fable 5.1 and Claude Mythos 5.1 on 1 September 2026. Both are the same model with different safeguard configurations; Mythos 5.1 is available only through a trusted access programme.
Primary sources used throughout:
All four rows below come from the comparison table on Anthropic’s launch page. Anthropic ran the table itself, including the competitor figures, which is stated in the video.
| Model | Score |
|---|---|
| Claude Fable 5.1 | 55.8% |
| Claude Opus 5 | 52.3% |
| Claude Fable 5 | 42.0% |
| GPT-5.6 Sol | 37.3% |
The relative improvement over Fable 5 is (55.8 − 42.0) / 42.0 = 32.9%, which the video states as a 33% relative jump.
Claude Mythos 5.1, under its more permissive cyber safeguards, reaches 60.9% on the same benchmark. The video does not use this figure.
Source: https://www.anthropic.com/claude-fable-and-mythos-5-1
| Model | Score |
|---|---|
| Claude Fable 5.1 | 52.6% |
| Claude Opus 5 | 29.0% |
| Claude Fable 5 | 24.7% |
| GPT-5.6 Sol | 22.4% |
Anthropic reports a standard error of roughly 3.5 to 4.5 points on this benchmark, and it is a young benchmark at version 0.1. Both caveats are stated in the video.
52.6 against 24.7 is a factor of 2.13; 52.6 against 22.4 is a factor of 2.35. The video’s “more than double” holds for both comparisons.
Source: https://www.anthropic.com/claude-fable-and-mythos-5-1
| Model | Score |
|---|---|
| Claude Fable 5.1 | 31.4% |
| Claude Opus 5 | 26.9% |
| GPT-5.6 Sol | 19.6% |
| Claude Fable 5 | 17.1% |
Source: https://www.anthropic.com/claude-fable-and-mythos-5-1
| Model | Score |
|---|---|
| Claude Fable 5.1 | 73.4% |
| Claude Fable 5 | 70.5% |
| Claude Opus 5 | 70.0% |
| GPT-5.6 Sol | 67.2% |
Fable 5.1 is 3.4 points above Opus 5 here (73.4 − 70.0), which the video contrasts with the 23.6 point gap on Terminal-Bench-Science.
Source: https://www.anthropic.com/claude-fable-and-mythos-5-1
| Model | Rating |
|---|---|
| Claude Fable 5.1 | 1853 |
| Claude Opus 5 | 1824 |
| Claude Fable 5 | 1723 |
| GPT-5.6 Sol | 1711 |
Source: https://www.anthropic.com/claude-fable-and-mythos-5-1
Claude Fable 5.1 is listed at $10 per million input tokens and $50 per million output tokens, unchanged from Claude Fable 5. Cache reads are $0.25 per million tokens, reduced from $1.00, a cut of 75%. Five minute cache writes are $12.50 per million and one hour cache writes are $20 per million. Batch API requests take a 50% discount.
Claude Opus 5 is listed at $5 per million input tokens and $25 per million output tokens, so Fable 5.1 is twice the headline price on both sides.
Source: https://platform.claude.com/docs/en/models/fable-5-1/overview
Anthropic estimates the cache read change makes typical Fable workloads cost about 25% less than Fable 5, and says savings on highly agentic work “will often be much larger, up to approximately 45%”. Anthropic states these figures are based on four weeks of its own August 2026 usage.
Source: https://www.anthropic.com/claude-fable-and-mythos-5-1
GPT-5.6 Sol is listed at $4 per million input tokens and $20 per million output tokens. OpenAI announced this as a promotional reduction in late August 2026, from $5 and $30, and has said it runs at least through 21 November 2026. Short context here means requests up to 272,000 input tokens.
Sources:
The Claude Fable 5.1 overview states: “For most workloads, start with Claude Opus 5. Use Claude Fable 5.1 for demanding reasoning and long-horizon agentic work, or when your evals on Claude Opus 5 at higher effort still fall short.”
Fable 5.1 is also listed with slower comparative latency than Opus 5, a 1M token context window, a 128K maximum output, and a reliable knowledge cutoff of June 2026.
Source: https://platform.claude.com/docs/en/models/fable-5-1/overview
Anthropic says Claude Code users can expect an average of around 60% fewer interventions per session from its cyber safeguards, compared with the safeguards that launched with Fable 5.
Fable 5.1 can now be used to identify software vulnerabilities in source code. Exploit generation, penetration testing and some binary based vulnerability scanning remain redirected or restricted.
In biology, Anthropic says the newer safeguards fire 85% less often on benign requests about elementary biology and medical questions than the safeguards that launched with Fable 5. Research and development life science work can still route to other models; Claude Mythos 5.1 offers the same capabilities to participants in Anthropic’s Project Glasswing trusted access programme, currently a set of United States organisations, with Anthropic saying it is coordinating with the US government to widen access.
Enterprise Frontier Safeguards let eligible customers store data on their own cloud infrastructure rather than Anthropic’s, with any human review done by the customer by default. Anthropic says it is available on Claude Code, Claude Enterprise, the Claude Platform, Amazon Bedrock and Microsoft Foundry, rolling out in phases from autumn 2026.
Sources:
These are figures the named organisations supplied to Anthropic for its launch. They are not independently reproduced benchmarks, which the video says.
Millennium. A piece of code had an extremely rare crash, about one run in a million, that nobody on the team had explained in four to five years. Anthropic says Fable 5.1 disassembled an external vendor library, matched it against the core dump, and traced the crash to a bug in that library.
MongoDB. Anthropic reports a complex prototype researched and built in about three days, including hours of unattended work and strong verification loops, with a full visual walkthrough.
Browserbase. 82% of tasks completed on its hardest browser agent benchmark, at about ten minutes each, against 74% for Claude Opus 5 and 57% for Claude Fable 5.
Crosby. RedlineBench, a contract redlining benchmark, rose from 47.9 to 57.0 over Fable 5.
Samaya. A 55.9% rubric score on its FrontierFinance benchmark, against 49.2% for Fable 5.
Source: https://www.anthropic.com/claude-fable-and-mythos-5-1