Public benchmarkExternal source reviewReviewed by DanPublished Source checked Updated
Fable 5.2: agentic coding benchmark source review
The report
Verdict
Published benchmark evidence
Two FrontierCode 2.0 publications report 49.1% and 50.2%, but their runs and operators differ. The small gap is not a measured model improvement or a local win.
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
- FrontierCode 2.0 — Anthropic: 49.1%. Fable 5.2; Xhigh effort; publisher evaluation. Reported by Anthropic vendor table.
- FrontierCode 2.0 — Cognition: 50.2%. Fable 5.2; xhigh; Claude Code harness; separate run. Reported by Cognition benchmark operator.
- FrontierSWE v2: 69.9%. Fable 5.2; agentic coding benchmark. Reported by Anthropic vendor table.
Original sources
Source 1 — FrontierCode 2.0 — Anthropic, FrontierSWE v2.
Source 2 — FrontierCode 2.0 — Cognition.
What this evidence tells a reader
FrontierCode 2.0 evaluates merge-worthy software changes, not simply whether code passes one test. The Anthropic-published 49.1% and Cognition-published 50.2% are separate evaluations; even for the same named suite, effort settings, harness details and run conditions can affect the outcome. Averaging the two or calling the 1.1-point gap a performance gain would hide those differences.
The 69.9% FrontierSWE v2 result covers a different coding assessment. It can add evidence that Fable was evaluated on agentic software work, but it cannot be ranked against a FrontierCode percentage. A high test score also leaves practical questions about reviewable diffs, unexpected file edits and the number of human corrections.
A useful next experiment is one real repository issue run with fixed instructions and a recorded harness, then compare the patch's acceptance, time, edits outside scope and review cost. This is a proposed reader method, not a test DanManREAL claims to have carried out for this page.
How to use these results
- Treat the two FrontierCode records as a range of published observations, not interchangeable trials.
- Prioritize a small code-maintenance trial with code review and tests before selecting a long-running agent.
- Check whether the chosen Claude Code harness and effort setting match the run behind the claimed score.
Limits of this report
- The two FrontierCode values were published by different operators under distinct runs; no common repeated-trial distribution is provided.
- The FrontierSWE result is a separate suite and says nothing by itself about local subscriptions, speed, or unexpected edits.
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 Fable 5.2
- FrontierCode 2.0 — Anthropic
- 49.1%Anthropic vendor table (model publisher). Fable 5.2; Xhigh effort; publisher evaluation. Not independently rerun by DanManREAL.
- FrontierCode 2.0 — Cognition
- 50.2%Cognition benchmark operator (independent operator or secondary publication). Fable 5.2; xhigh; Claude Code harness; separate run. Not independently rerun by DanManREAL.
- FrontierSWE v2
- 69.9%Anthropic vendor table (model publisher). Fable 5.2; agentic coding benchmark. 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-fable-5-2-public-evidence
Models covered by external benchmarks
On the tier list's bench
- Claude Fable 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