Public benchmarkExternal source reviewReviewed by DanPublished Source checked Updated
Muse Wonder Full 1.0: public coding benchmark receipts
The report
Verdict
Published benchmark evidence
Meta's 79.0% DeepSWE, 57.5% SWE-Marathon and 57.1% AA Terminal-Bench results are for Muse Wonder Full 1.0 under different published conditions.
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
- DeepSWE v1.1: 79.0%. Muse Wonder Full 1.0; max effort. Reported by Meta vendor model card.
- SWE-Marathon: 57.5%. Muse Wonder Full 1.0; xhigh effort. Reported by Meta vendor model card.
- Terminal-Bench 4.0 (AA): 57.1%. Muse Wonder Full 1.0; Artificial Analysis track. Reported by Meta vendor model card.
Original sources
Source 1 — DeepSWE v1.1, SWE-Marathon, Terminal-Bench 4.0 (AA).
What this evidence tells a reader
The model name matters here: Full 1.0 is not a generic label for every Muse Wonder deployment. A reader who obtains the base variant or a differently configured host cannot assume it inherits the Full variant's published results. Check the exact endpoint and model version before making a selection.
DeepSWE v1.1 at 79.0% and SWE-Marathon at 57.5% are two types of software engineering evaluation, not stages of the same test. Terminal-Bench 4.0 at 57.1% explicitly names the Artificial Analysis track, which limits direct comparison with non-AA terminal figures elsewhere on the site.
For a practical coding assignment, the useful next step is to run the candidate through the actual editor and agent tools, then inspect diffs, retries, elapsed time and any breaking changes. A high publisher score does not verify that the model is available through DanManREAL's Muse Code route or that it will deliver an acceptable artifact.
How to use these results
- Confirm 'Full 1.0' rather than a base or renamed variant.
- Treat the AA terminal result as track-qualified, not a universal terminal success rate.
- Test one representative multi-file patch in the target harness before choosing an implementation model.
Limits of this report
- All three scores come from Meta's model-card reporting, not a reproduced DanManREAL run.
- The public profile does not establish actual Muse Code behavior, token cost or review effort for a specific project.
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
Muse Wonder Full 1.0
- DeepSWE v1.1
- 79.0%Meta vendor model card (model publisher). Muse Wonder Full 1.0; max effort. Not independently rerun by DanManREAL.
- SWE-Marathon
- 57.5%Meta vendor model card (model publisher). Muse Wonder Full 1.0; xhigh effort. Not independently rerun by DanManREAL.
- Terminal-Bench 4.0 (AA)
- 57.1%Meta vendor model card (model publisher). Muse Wonder Full 1.0; Artificial Analysis track. 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/muse-wonder-full-1-0-public-benchmarks
Models covered by external benchmarks
On the tier list's bench
- Muse Wonder Full 1.0 · 2 receipts 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