Quick answer: GPT-6.1 Sol is OpenAI's updated mid-tier model, announced September 29, 2026 as an incremental refresh of GPT-6 Sol. It's priced at $2 per million input tokens and $10 per million output tokens ($0.10 for cache reads), available via the OpenAI API as
gpt-6.1-soland through ChatGPT Work and Codex — not yet rolled out to regular ChatGPT Plus/consumer tiers.
What it is: An incremental update to GPT-6 Sol, OpenAI's faster, cheaper complement to the flagship GPT-6 Astra line, with gains on agentic coding benchmarks reported relative to GPT-6 Sol.
Where it's strong: OpenAI reports gains over GPT-6 Sol on agentic software-engineering tasks (DeepSWE, AutomationBench), described in relative terms rather than exact absolute figures in all cases.
Availability: API (model name gpt-6-1-sol) plus ChatGPT Work and Codex; not yet available on regular ChatGPT consumer tiers at launch.
GPT-6.1 Sol is OpenAI's refreshed mid-tier model, announced September 29, 2026 as a successor to GPT-6 Sol within the same product tier (a faster, cheaper complement to the flagship GPT-6 Astra line, mirroring how GPT-5.6 Sol related to earlier GPT-5-generation Astra models). OpenAI's announcement emphasizes gains on agentic coding and software-engineering benchmarks relative to GPT-6 Sol, in some cases reported as relative deltas (e.g. "within 2.1 points of GPT-6 Astra on OSWorld 2.0") rather than fully independent absolute scores for every benchmark cited.
Pricing is unchanged from GPT-6 Sol's tier: $2 per million input tokens, $10 per million output tokens, and $0.10 per million tokens for cache reads. At launch, GPT-6.1 Sol is available via the OpenAI API (as gpt-6.1-sol) and through ChatGPT Work and Codex, but has not yet rolled out to regular ChatGPT Plus/consumer surfaces.
| Field | Value |
|---|---|
| Organization | OpenAI |
| API model name | gpt-6-1-sol |
| License | Proprietary (API only) |
| Modality | Multimodal (text + vision) |
| Release date | 2026-09-29 |
| Availability | OpenAI API, ChatGPT Work, Codex |
| Price per 1M tokens | GPT-6.1 Sol |
|---|---|
| Input tokens | $2.00 |
| Output tokens | $10.00 |
| Cache reads | $0.10 |
| Benchmark | Score | Source | Date |
|---|---|---|---|
| DeepSWE (v1.1) | ≈75.2% (derived) | OpenAI: GPT-6.1 Sol | 2026-09 |
| AutomationBench | ≈42.2% (derived) | OpenAI: GPT-6.1 Sol | 2026-09 |
The DeepSWE and AutomationBench figures above are derived from relative-gain statements in OpenAI's GPT-6.1 Sol announcement (improvement over GPT-6 Sol's published scores of ~68.8% DeepSWE and ~41.4%-adjacent AutomationBench baselines) rather than stated as precise absolute scores — treat as approximate pending a more precise primary disclosure. OpenAI also reported GPT-6.1 Sol performing "within 2.1 percentage points" of GPT-6 Astra on OSWorld 2.0, without giving Sol's exact score, so that benchmark is not included in the table above.