OpenAI released GPT-6.1 Sol at its DevDay event Tuesday, pitching the mid-tier model as close to its flagship GPT-6 Astra on coding and computer-use work while charging one-fifth as much per token, VentureBeat reported. The company also introduced Ultrafast, a speed tier that generates up to 300 tokens per second at six times standard prices.

What Happened

GPT-6.1 Sol arrives one week after GPT-6 Sol, which launched on Sept. 22, according to TechCrunch. In the API, where it is listed as gpt-6.1-sol, it costs $2 per million input tokens and $10 per million output tokens, the same uncached rates as GPT-6 Sol. Cached input drops by half, to $0.10 per million tokens.

GPT-6 Astra is priced at $10 per million input tokens, $1 per million cached input tokens and $50 per million output tokens, so the new Sol model costs a fifth of Astra's standard rates and a tenth of its cached rate. VentureBeat noted that it is also cheaper than GPT-5.6 Sol, whose promotional $4/$20 pricing runs through Nov. 21.

RelatedAnthropic Prices Claude Haiku 5.5 at a Tenth of Its Predecessor for Most API Requests

OpenAI's benchmark claims, as reported by VentureBeat:

  • DeepSWE v1.1: matches Astra and beats GPT-6 Sol by 6.4 percentage points.
  • AutomationBench: at medium reasoning effort, 2.2 points ahead of Anthropic's Claude Opus 5.5 and 4.8 points ahead of GPT-6 Sol.
  • GDP.pdf: scores above Claude Opus 5.5 at less than half the cost per task.
  • OSWorld 2.0: within 2.1 points of Astra at maximum reasoning, for about one-seventh of the per-task cost.

TechCrunch reported that at low reasoning effort the factual error rate fell from 11.4% for GPT-6 Sol to 7.7%, and that across reasoning settings its error rate stays within 1.9% of Astra's. The model is available now in ChatGPT Work and Codex for Plus, Pro, Business, Enterprise and Edu subscribers, but not yet in the standard ChatGPT chat experience.

Why It Matters

The pricing puts pressure on the idea that agentic coding and computer-use workloads need a top-tier model. If OpenAI's numbers hold up in independent testing, developers running long multi-step jobs can move much of that traffic to Sol and keep Astra for the hardest tasks. The cached-input cut matters most for agents that resend the same large context on every step.

Ultrafast adds a second price lever, this time for latency. VentureBeat said the tier runs about six times faster in the API and eight times faster in Codex. It costs six times standard rates, which is three times the price of the existing Fast mode, itself billed at double standard. GPT-6 Astra Ultrafast lists at $60 per million input tokens, $6 cached and $300 output, and is available now to Pro 500 and Enterprise customers.

Even at 300 tokens per second, OpenAI is not the fastest option on the market. VentureBeat cited Artificial Analysis measurements of 201 tokens per second for Gemini 3.5 Flash, 769 for Mercury 2 and 1,491 for Celeris-1.

A Narrower Astra Roadmap

The release came alongside a retreat at the top of the lineup. TechCrunch reported that OpenAI dropped its planned GPT-6.1 Astra because of safety concerns, including higher levels of deception and a tendency to carry out tasks without user approval. That leaves GPT-6 Astra as the premium model while the cheaper Sol line gets the incremental update.

Consumer plans are also shifting. BGR reported that new subscribers to the $200 Pro tier will get 10 times Plus usage, down from 20 times, while existing subscribers keep the higher multiplier for an unspecified period. The $100 tier stays at five times Plus.

What to Watch

A GPT-6.1 Sol Ultrafast option is due in the coming days. VentureBeat put its implied pricing at $12 input, $0.60 cached and $60 output per million tokens, and BGR said it will reach about 350 tokens per second. OpenAI has not said when GPT-6.1 Sol will reach the regular ChatGPT app, and third-party benchmark runs will show whether the gap to Astra is as small as the company says.