Public benchmarkExternal source reviewReviewed by DanPublished Source checked Updated
Mythos 5.2: research and security benchmark source review
The report
Verdict
Published benchmark evidence
The strongest published evidence here concerns specialized science and security evaluations, not ordinary app development. Anthropic published these results; DanManREAL did not rerun them.
On the tier list: The November 17 tier is Dan’s own ranking; public benchmark numbers are a separate evidence channel, not an automatic tier adjustment.
What the benchmarks report
- Terminal-Bench-Science 0.1: 75.3%. Mythos 5.2; agentic scientific terminal research. Reported by Anthropic vendor table.
- CyScenarioBench: 76.3%. Mythos 5.2; cybersecurity scenarios. Reported by Anthropic vendor table.
- CVE Rediscovery: 89.6%. Mythos 5.2; security-research evaluation. Reported by Anthropic vendor table.
Original sources
Source 1 — Terminal-Bench-Science 0.1, CyScenarioBench, CVE Rediscovery.
What this evidence tells a reader
The 75.3% science-terminal figure concerns research-oriented agent work. The 76.3% cyber-scenario figure and 89.6% CVE-rediscovery figure concern different security tasks. Reading 89.6% as a general probability of finding a vulnerability would be a mistake: the benchmark's selected problems, success definition and access conditions determine what the number measures.
Access and safeguards matter as much as the result. Mythos and Fable share a model lineage but are offered for different work with different safety controls. A general coding benchmark from Fable cannot substitute for this specialist evidence, and a security score cannot establish that an assistant may safely take autonomous actions on production infrastructure.
For an authorized security-research workflow, these figures can justify evaluating Mythos on a small set of permitted, reproducible research tasks. They do not justify granting broad machine permissions or publishing vulnerabilities without independent verification.
How to use these results
- Use this report when scoping specialist science or authorized cybersecurity tasks; check the actual access conditions first.
- Before adoption, reproduce a permitted task and record correctness, false leads, review effort and permission boundaries.
- For normal coding decisions, consult separate code-maintenance evidence rather than treating security scores as a coding leaderboard.
Limits of this report
- All three published percentages come from an Anthropic table; an independent replication with the same dataset and safeguards is not supplied here.
- This source review cannot establish actual access eligibility, false-positive rates on a user's systems, or safe autonomous deployment behavior.
Follow the lab
New test reports and tier-list changes get announced in my Discord's #news channel. The full videos are on YouTube.
Receipts
DanManREAL · AI lab
Test receipt
Claude Mythos 5.2
- Terminal-Bench-Science 0.1
- 75.3%Anthropic vendor table (model publisher). Mythos 5.2; agentic scientific terminal research. Not independently rerun by DanManREAL.
- CyScenarioBench
- 76.3%Anthropic vendor table (model publisher). Mythos 5.2; cybersecurity scenarios. Not independently rerun by DanManREAL.
- CVE Rediscovery
- 89.6%Anthropic vendor table (model publisher). Mythos 5.2; security-research evaluation. Not independently rerun by DanManREAL.
- Source review
- November 17, 2026Date these source claims were reviewed; not a claim about when the published benchmark was executed.
Verdict
Published benchmark evidence
Tier list The November 17 tier is Dan’s own ranking; public benchmark numbers are a separate evidence channel, not an automatic tier adjustment.
danmanreal.com/tests/claude-mythos-5-2-public-evidence
Models covered by external benchmarks
On the tier list's bench
- Claude Mythos 5.2 · 1 receipt on file
Related
Independent testing. Sponsorships are always disclosed, and verdicts are never for sale.
Disagree? Bring your receiptsAll reports
- One inbox for my AI coding agents: triage took 34% less time, resuming work did not improveHead-to-headNov 20
- Mythos 5.2: research and security benchmark source reviewPublic benchmarkNov 17
- Fable 5.2: agentic coding benchmark source reviewPublic benchmarkNov 17
- Claude Opus 5.5: public coding benchmark receiptsPublic benchmarkNov 17
- Astra 6.2: new model and public task-evaluation evidencePublic benchmarkNov 17
- Sol 6.2: new model and public coding benchmark evidencePublic benchmarkNov 17
- Gemini 4 Carbon: benchmark source and usage profilePublic benchmarkNov 17
- Qwen 4: Qwen4-Max launch and benchmark sourcesPublic benchmarkNov 17
- Kimi K3.1 Pro: public coding benchmark receiptsPublic benchmarkNov 17
- GLM-5.3: public coding benchmark receiptsPublic benchmarkNov 17
- Gemini 4 Argon: public coding benchmark receiptsPublic benchmarkNov 17
- Claude Sonnet 5.5: public coding benchmark receiptsPublic benchmarkNov 17
- DeepSeek V4.1 Flash: public coding benchmark receiptsPublic benchmarkNov 17
- Muse Wonder Full 1.0: public coding benchmark receiptsPublic benchmarkNov 17
- Grok 4.7: separate release and public coding resultsPublic benchmarkNov 17
- ai-design-check on two of my FABLE screens: 10 warnings, 4 of them false alarmsField testNov 2
- Vercel Agent Desktop used a sample Trip app in 6 actions without moving the mouseField testNov 1
- Cerberus Guard on sample data: one crawler name, two verdicts, and a block earned by behaviorField testOct 31
- GPT-6.1 Terra vs Opus 5.5: three real projects, and the designs I picked blindHead-to-headOct 29
- Steps → Reduced: opening a terminal agent went from 98 seconds to 11AuditOct 28
- One ChatGPT chat directed Opus and Codex at once through DMR_REMOTEField testOct 28
- Claude Duet ran my channel for a week and caught 269 fake AI crawlersField testOct 27
- Kimi K3.1 Pro built my animation family, after a quota stop and 21 minutes of failed requestsField testOct 24
- ChatGPT + Codex built 33 new DMR_REMOTE toolsField testOct 23
- Gemini 4 Argon reviewed a finished episode: 14 clips, one flag, one inconsistencyField testOct 21
- Qwen 3.8 Flash Next on my RTX 5080: 81 tokens a second with 202K tokens readAuditOct 21
- Does an extra AI review step pay off? A small DeepSeek trial said noHead-to-headOct 19
- My real DeepSeek bill: 12.31 billion tokens for $122.28Cost receiptOct 17
- Gemini 4 Argon kept hanging in Gemini CLI. The cause was a 60-second timeoutIncidentOct 14
- I asked 10 AI coding agents for their biggest weaknessSelf-reportOct 14
- Jarvis 27B on my own PC: 30 tokens a second, and it still recalls at 251KAuditOct 10
- My agents' first game benchmark was wrong. Here's how it was caughtIncidentOct 10
- Pixal3D + Blender turned my 2D art into animated 3D charactersField testSep 22
- DeepSeek Flash answered 200 OK and sent nothing for two hoursIncidentSep 14
- Grok, Codex and DeepSeek played WorldOS for Rank 1. Grok won with 98Head-to-headAug 20
- DeepSeek V4 Flash audited 18 of my real coding sessions. Two said 'done' too earlyAuditAug 4