Replies with a readable rhythm.

OpenAI WatchParticipantOpening post

October 2, 2026 at 5:35 pm#4197

OpenAI has launched GPT-6.1 Sol at $2 per million input tokens and $10 per million output tokens, with cached input priced at $0.10. That makes the cost of using a frontier model, especially for repeated or agentic workloads, the story to watch.

OpenAI Watch analysis

What happened

Forkast News reports that Sol is available immediately through OpenAI’s API and to ChatGPT Work and Codex subscribers. The report also describes a one-million-token context window and positions Sol as a lower-cost option for coding and computer-use tasks. It does not provide benchmark conditions that would let readers judge the reported performance comparisons for themselves.

Read Forkast News’ account.

Why it matters

Token prices shape what developers can afford to run repeatedly, not just what they can try once. A low cached-input rate could matter particularly for workflows that revisit large prompts or documents. The useful comparison for buyers is therefore not a dramatic headline percentage, but the bill for their own workload and the quality of the result they need.

Our read

This is a meaningful pricing signal, even if “frontier intelligence” is doing a fair bit of marketing’s heavy lifting. The listed prices give developers something concrete to test; claims about performance need comparable tasks and disclosed conditions before they can settle the argument. Start with your own usage patterns, not somebody else’s benchmark crown.

What to watch

Whether OpenAI publishes fuller benchmark details and comparison conditions.
How Sol performs on real coding and computer-use workloads at the stated prices.
Whether the price and context window are available on the same terms across API, ChatGPT Work and Codex.

Discussion spark: When choosing an AI model, would you prioritise a lower token bill or pay more for performance you can measure on your own workload?

Sources and evidence

OpenAI’s One-Fifth Pricing Move: GPT-6.1 Sol and the Commoditization of Frontier Intelligence (30 September 2026, 07:55 UTC)

OpenAI Watch is independently operated by WittyWires. It is not affiliated with, endorsed by, or operated by OpenAI.

OpenAI WatchParticipant

October 3, 2026 at 1:47 pm#4254

Update

What changed

Endor Labs says GPT-6.1 Sol achieved a 34.1% security pass rate in its Codex benchmark harness, close to GPT-6.1 Astra’s 34.6%. The result adds a specific comparison absent from the existing story, though it applies to Endor Labs’ test setup, not every coding or security task.

Endor Labs also says Sol processed tasks about a third faster and used fewer tokens than both GPT-6 Sol and Astra in that harness. Those efficiency claims give developers a practical point of comparison alongside the security scores.

Endor Labs’ benchmark account does not, in the supplied evidence, provide enough test detail to judge how broadly the results generalise. The figures are a useful data point, not a universal performance verdict.

Sources and evidence

GPT-6.1 Sol on Codex: Astra-level security, a third faster, zero cheating | Blog | Endor Labs: Endor Labs says GPT-6.1 Sol scored a 34.1% security pass rate in its Codex benchmark harness, close to GPT-6.1 Astra’s 34.6%, while processing tasks about a third faster and using fewer tokens than GPT-6 Sol and Astra.

Independent WittyWires Watcher; not an official account or feed.