gpt.college

OpenAI · GPT-6.1 · Reviewed 2026-10-02

GPT-6.1 Sol

OpenAI's updated Sol model for coding, computer use, and professional work at lower cost than Astra.

All models
Overall · 0–10088.6
Status
active
Released
2026-09-29
Context
1,050,000 tokens
Parameters
Undisclosed
Inputs
Text · Image
Type
multimodal
Max output
128,000 tokens
API model ID
gpt-6.1-sol
Reasoning
low / medium / high / xhigh / max
Knowledge cutoff
2026-04-30
Training cutoff
—
Weights / license
Unavailable
Price / 1M tokens
$2 in · $10 out
Cached input
$0.1
Complex codingComputer useLower cached-input price
CodingReasoningResearchMultimodalAgentic

Independent benchmark scores

Independently evaluated

Reasoning

53%

Humanity's Last Exam as evaluated by Artificial Analysis; the source configuration is preserved.

Artificial Analysis · max

Independently evaluated

Research

1575 Elo

GDPval-AA v2.1 professional knowledge-work Elo. This is a task lens, not an assessment of every research skill.

Artificial Analysis · max

Independently evaluated

Multimodal

31%

GDP.pdf professional document understanding, including visual tables, charts and diagrams; not an audio/video score.

Artificial Analysis · max

Independently evaluated

Agentic

65%

AutomationBench-AA multi-step business workflows in the Artificial Analysis harness.

Artificial Analysis · max

Raw benchmark results

Source-backed rows with the original evaluation context preserved.

9 results
BenchmarkScoreSourceEvaluation notes
AutomationBench-AAChecked 2026-10-02 65% Artificial Analysis Artificial Analysis Intelligence Index v4.3.2; configuration=max. Rounded public score; checked date is not the evaluation execution date.
AA GDP.pdfChecked 2026-10-02 31% Artificial Analysis Artificial Analysis Intelligence Index v4.3.2; configuration=max. Rounded public score; checked date is not the evaluation execution date.
GDPval-AA v2.1Checked 2026-10-02 1575 Artificial Analysis Artificial Analysis Intelligence Index v4.3.2; configuration=max. Rounded public score; checked date is not the evaluation execution date.
AA Humanity's Last ExamChecked 2026-10-02 53% Artificial Analysis Artificial Analysis Intelligence Index v4.3.2; configuration=max. Rounded public score; checked date is not the evaluation execution date.
Artificial Analysis Intelligence Index v4.3.2Checked 2026-10-02 52 Artificial Analysis Artificial Analysis Intelligence Index v4.3.2; configuration=max. Rounded public score; checked date is not the evaluation execution date.
AA Terminal-Bench 4.0Checked 2026-10-02 56% Artificial Analysis Artificial Analysis Intelligence Index v4.3.2; configuration=max. Rounded public score; checked date is not the evaluation execution date.
AutomationBench v1.0.6Checked 2026-10-02 31.7% OpenAI Developer Community Community transcription; reasoning=medium. Official release identifies AutomationBench 1.0.6.
DeepSWE v1.1Checked 2026-10-02 75.2% OpenAI Developer Community Community transcription of OpenAI release chart; reasoning=high. Official release confirms benchmark context, not this numerical transcription.
OSWorld 2.0 Offline PartialChecked 2026-10-02 71.4% OpenAI Developer Community Community transcription; reasoning=max. Official release specifies offline partial reward, v2026.08.08; not latency-simulated OSWorld.

Strengths

  • Complex coding
  • Computer use
  • Lower cached-input price

Limitations

  • Benchmark effort and offline task sets affect comparability
  • Independent AA v4.3.2 scores are now available; older community chart transcriptions remain separately labeled

Model evidence · official · Checked 2026-10-02

Introducing GPT-6.1 Sol

Release and evaluation context. OSWorld uses partial reward on the offline v2026.08.08 task set; numerical chart transcriptions have a separate community source.

Model evidence · official · Checked 2026-10-02

GPT-6.1 Sol model specifications

Official standard API limits, reasoning settings and pricing including the 272K input threshold.

Model evidence · community · Checked 2026-10-02

GPT-6.1 Sol cost-performance chart transcription

Community announcement transcribes 75.2 DeepSWE/high, 31.7 AutomationBench/medium and 71.4 OSWorld offline/max. Traceable but not an independent evaluation or an official numerical table.

Task lenses preserve raw scores. Radar axes use score / current leader × 100; overall is the equal-weight average of all five indices. Missing evidence is not a zero or a low rank.