GPT-6.1 Sol Lands One Point Behind GPT-6 Astra on Artificial Analysis Index
A.I. / news
GPT-6.1 Sol Lands One Point Behind GPT-6 Astra on Artificial Analysis Index
OpenAI prices Sol at one-fifth of Astra. The independent tracker also logged a 168-second wait for the first token at maximum effort.
OpenAI's GPT-6.1 Sol scored 52 on the Artificial Analysis Intelligence Index v4.3.2 at maximum effort, one point behind GPT-6 Astra at 53, according to OfficeChai's report on Sept. 30. OpenAI released Sol on Sept. 29.
The original GPT-6 Sol scored 48 on the same index, so the update adds four points. OpenAI lists Sol at $2 per million input tokens and $10 per million output tokens, one-fifth of Astra's $10 and $50.
Where GPT-6.1 Sol ranks on the Intelligence Index
OfficeChai reported Anthropic's Claude Opus 5.5 at 58 and Claude Sonnet 5.5 at 56. Sol's 52 puts it ahead of Claude Opus 5 at 51 and behind Claude Fable 5.1 at 53.
- Claude Opus 5.558 points
- Claude Sonnet 5.556 points
- GPT-6 Astra53 points
- Claude Fable 5.153 points
- GPT-6.1 Sol52 points
- Claude Opus 551 points
- GPT-6 Sol48 points
Source: OfficeChai citing Artificial Analysis, Sept. 30, 2026
Artificial Analysis's own page for the xhigh setting shows 51, ranking 14th of 224 models, with a median of 26. The one-point gap between 51 and 52 is a difference in effort setting, which OfficeChai labelled max.
Speed and cost: 49.8 tokens a second
The same Artificial Analysis page logged an output speed of 49.8 tokens per second for the xhigh setting, ranked 150th. Time to first token was 168.69 seconds. The page calls the model "notably slow, however fairly concise", and says it generated 36 million output tokens during the index run, against a median of 81 million.
The average cost per task on the index was $0.39. OfficeChai said Artificial Analysis had not published a Sol cost per task when it wrote, and put GPT-6 Sol at about $1.06.
OpenAI's own numbers, and the Ultrafast tier
VentureBeat reported, from OpenAI's launch materials, that Sol matches Astra on DeepSWE v1.1 and beats GPT-6 Sol by 6.4 percentage points. It beats Claude Opus 5.5 on AutomationBench by 2.2 points at medium reasoning. On OSWorld 2.0 it is within 2.1 points of Astra at maximum effort. These are vendor figures, and VentureBeat said OpenAI's evaluations can differ from production ChatGPT.
Sol is available through the API and in ChatGPT Work and Codex, not in regular ChatGPT. An Ultrafast tier of up to 300 tokens per second is live for Astra and coming for Sol.
| Tier | Input per 1M | Output per 1M |
|---|---|---|
| GPT-6 Astra | $10 | $50 |
| GPT-6.1 Sol | $2 | $10 |
| GPT-6 Astra Ultrafast | $60 | $300 |
| GPT-6.1 Sol Ultrafast, projected by VentureBeat | $12 | $60 |
VentureBeat cited Artificial Analysis for Mercury 2 at about 769 tokens per second and Celeris-1 at about 1,491, both faster than Ultrafast.
Cost has been the theme of other OpenAI coverage here, from the Synopsys revenue-share deal to the AGMAI math results.
OpenAI has not given a date for Sol Ultrafast beyond VentureBeat's report that it is coming in days.
Sources
More in A.I.
- 01GPT-6.1 Sol Is Priced at One-Fifth of Astra, and Its System Card Rates Cyber CriticalOpenAI's Sept. 29 addendum also shows the model misrepresenting its own coding work more often than GPT-6 Astra did.
- 02Gemini 4 Argon Goes to Cyber Defenders First, With Broad Access UndatedGoogle priced its new frontier model at $2 and $10 per million tokens for an introductory period, then $4 and $20, and has not said when most developers get it.
- 03Aleph Alpha's Kolibri Ships Under Apache 2.0, Compared Only With Spring ModelsThe 78B-parameter German-English model activates 3.46B per token, and its published benchmark table leaves out every open-weight release since the spring.
- 04Runway's Praxis-1 Robot Model Is Open-Weight on Paper, With Weights Still UnreleasedThe video-trained control model is being tested by Noble Machines, Standard Bots and Ultra, and Runway has not published a parameter count or a licence.