The Trace Manifest
Subscribe
Corrections Subscribe

GPT-5.6 Sol

OpenAI · Closed · 5.6

Flagship OpenAI coding and agent model. Strongest on broad agentic coding benchmarks.

Status
active
Release date
2026-07-14
Parameters
unknown
Context window
128K
Modalities
text,code,image
Capabilities recorded
tool use, structured output, API
Last verified
2026-07-20

Reviewed use cases

  • agentic coding
  • terminal use
  • complex reasoning

Recorded limitations

  • SWE-Bench Pro lags behind Claude Fable 5

Benchmark Performance

77 / 100   TRACE Aggregate Score

TRACE Score 77/100 based on 4 benchmarks. 1 independent run (weighted 3×). 3 vendor-reported runs (weighted 1×). Last updated 01/07/2026.

BenchmarkScoreProvenanceDateSource
SWE-bench Verified 72.7% Vendor 01/07/2026 Source
LMSYS Chatbot Arena ~1350 Independent 01/07/2026 Source
MMLU-Pro 88.0% Vendor 01/07/2026 Source
HumanEval+ 95.0% Vendor 01/07/2026 Source

Pricing is omitted until provider-model pricing rows have their own completed review and publication state.