GPT-6.1 Sol vs Muse Wonder Full
AutomationBench-AA AutomationBench-AA: Agents orchestrate SaaS workflows through APIs; objectives receive credit only when the task has no guardrail violation
GPT-6.1 Sol62.6
Muse Wonder Full71.3
GDPval-AA (work tasks) 2.1 GDPval-AA (work tasks) 2.1: Real work tasks from many professions (reports, spreadsheets, analyses), graded against expert work
GPT-6.1 Sol47.5
Muse Wonder Full67.9
Terminal-Bench 4.0 Terminal-Bench 4.0 · source setting 2: Agents use a terminal to complete executable coding, system administration and research tasks in the 4.0 suite
GPT-6.1 Sol48
Muse Wonder Full57.1
AA-Briefcase 1.1 · Elo AA-Briefcase 1.1 · Elo: Agents produce professional deliverables; blind pairwise comparisons produce the source Elo rating
GPT-6.1 Sol1,361.2
Muse Wonder Full1,729.3
AA-LCR (long context) 1.1 AA-LCR (long context) 1.1: Questions that need reading and connecting facts across very long documents (around 100K tokens)
GPT-6.1 Sol83.3
Muse Wonder Full83.7
Harvey LAB-AA 1.1 Harvey LAB-AA 1.1: Legal agents work from case documents; every rubric criterion must pass and material hallucinations receive no credit
GPT-6.1 Sol · max6.9
Muse Wonder Full13.1
AA Analyst Agent AA Analyst Agent: Data-analyst tasks: the model explores data with tools and reports findings
GPT-6.1 Sol · max50
Muse Wonder Full—
Terminal-Bench Science 0.1 Terminal-Bench Science 0.1: Agents complete expert-authored research workflows across five scientific domains in a terminal
GPT-6.1 Sol · max58.1
Muse Wonder Full—
DeepSWE 1.1 DeepSWE 1.1: Agents change repository code to resolve software engineering tasks under the stated DeepSWE edition
GPT-6.1 Sol · max72.3
Muse Wonder Full—
FrontierCode 1.1 FrontierCode 1.1: Agents implement changes in real repositories, graded against maintainer-defined requirements
GPT-6.1 Sol · max49.1
Muse Wonder Full—
SWE-bench Pro (v2 full) SWE-bench Pro (v2 full): Agents resolve repository issues under the network-locked v2 protocol; full and 51-task Hard results are separate
GPT-6.1 Sol · xhigh96.1
Muse Wonder Full—
EBR Bench EBR Bench: Agents learn to play Earthborne Rangers across repeated playthroughs
GPT-6.1 Sol · max54.3
Muse Wonder Full—
Furniture Assembly Furniture Assembly: Models inspect furniture assembly photos, identify incorrect builds and locate the mistaken step
GPT-6.1 Sol · max80
Muse Wonder Full—
SpireClimb SpireClimb: Agents play Slay the Spire under the named game and difficulty setting
GPT-6.1 Sol · max24.7
Muse Wonder Full—
SpireClimb StS1 A20 Heart SpireClimb StS1 A20 Heart: Agents play Slay the Spire under the named game and difficulty setting
GPT-6.1 Sol · max20
Muse Wonder Full—
SpireClimb StS2 A10 SpireClimb StS2 A10: Agents play Slay the Spire under the named game and difficulty setting
GPT-6.1 Sol · max29.3
Muse Wonder Full—
| Benchmark | GPT-6.1 Sol | Muse Wonder Full |
|---|---|---|
| AutomationBench-AA | 62.6 | 71.3 |
| GDPval-AA (work tasks) 2.1 | 47.5 | 67.9 |
| Terminal-Bench 4.0 | 48 | 57.1 |
| AA-Briefcase 1.1 · Elo | 1,361.2 | 1,729.3 |
| AA-LCR (long context) 1.1 | 83.3 | 83.7 |
| Harvey LAB-AA 1.1 | 6.9 · max | 13.1 |
| AA Analyst Agent | 50 · max | — |
| Terminal-Bench Science 0.1 | 58.1 · max | — |
| DeepSWE 1.1 | 72.3 · max | — |
| FrontierCode 1.1 | 49.1 · max | — |
| SWE-bench Pro (v2 full) | 96.1 · xhigh | — |
| EBR Bench | 54.3 · max | — |
| Furniture Assembly | 80 · max | — |
| SpireClimb | 24.7 · max | — |
| SpireClimb StS1 A20 Heart | 20 · max | — |
| SpireClimb StS2 A10 | 29.3 · max | — |
Specs
| Spec | GPT-6.1 Sol | Muse Wonder Full |
|---|---|---|
| Input, per 1M | $2 | — |
| Output, per 1M | $10 | — |
| Cost per task? | $0.21 | — |
| Speed, tok/s | 45 | — |
| Context | 1.05M | 1M |
| Weights | Closed | Closed |
| Released | Sep 2026 | Oct 2026 |
| See images | ✓ | ✓ |
| Read PDFs | Partial | ? |
| Hear audio | — | ✓ |
| Watch video | — | ✓ |
| Call tools | ✓ | ✓ |
| Force a tool call | Partial | ? |
| Return strict JSON | ✓ | ? |
| Search the web | ✓ | ? |
| Run code | ✓ | ? |
| Use a computer | ✓ | ? |
What breaks when you switch
GPT-6.1 Sol → Muse Wonder Full
Context: 1.05M → 1M
Only in Muse Code›
Muse Wonder Full requires a Muse Code subscription; max reasoning effort on the Model API Wonder model does not select Full.
Muse Wonder Full → GPT-6.1 Sol
Remove unsupported parameters›
With reasoning enabled, remove temperature, top_p and top_logprobs, plus Chat Completions logprobs or Responses message.output_text.logprobs includes.
Responses API for tool calling · Chat Completions›
Use Responses for tools; Chat Completions supports this model without tool calling.
No none or minimal reasoning›
The none and minimal reasoning efforts are unsupported; choose low, medium, high, xhigh or max.
More than 272K input tokens›
Prompts above 272K input tokens charge 2× input and cache rates and 1.5× output rates for the entire request.
Cache writes billed at 1.25x›
Cache writes cost 1.25× the uncached input token rate.
Ultrafast prices 6x Standard›
Ultrafast mode costs 6× Standard token prices.