OpenAI releases GPT-6 Sol with 7.7% improvement and half API cost
OpenAI released GPT-6 Sol (Max) on September 22, 2026, reporting a +7.7% net improvement across more than 4,000 real‑world agentic sessions on the Agent Arena leaderboard. The new model sits at 6, up from 8 for GPT‑5.6 Sol (xHigh). API pricing drops to $2 per million input tokens and $10 per million output tokens, a 50% reduction from GPT‑5.6 rates.
Key points
- GPT‑6 Sol achieves +7.7% net improvement on Agent Arena, up from GPT‑5.6 Sol
- API pricing halves to $2/m input and $10/m output, a 50% reduction
- DeepSWE v1.1 score 68.8%, 80% cheaper than top Claude models
On the Confirmed Success metric GPT‑6 Sol scored +11.4% and ranked #4 overall, showing stronger reliability for completed tasks. On DeepSWE v1.1 it posted 68.8%, trailing Anthropic’s Claude Fable 5 but at about 80% lower cost. OpenAI also launched GPT‑6 Luna the same day, targeting scalable, cost‑effective agentic workflows, automation pipelines and always‑on AI infrastructure.
The pricing war continues with Anthropic’s Claude Opus 5 and Claude Fable 5 maintaining higher DeepSWE scores, while AutomationBench results suggest GPT‑6 Sol completes tasks at lower cost than some Claude Opus 5 configurations.
Model pages: GPT-6 Sol (Max) → · GPT-6 Luna → · GPT-6 Astra →
The story so far
4 episodes →- OpenAI releases GPT-6 Sol with 7.7% improvement and half API costthis story
OpenAI’s GPT-6 Sol (Max) achieves 7.7% net improvement while cutting API costs in half
cryptobriefing.com · 25 September 2026
Loading the full article…
This text was published by cryptobriefing.com and written by Diego Almada Lopez. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗
Coverage and discussion
3sourcesThe headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Generative AI & Models
All →- Anthropic releases Claude Opus 5.5 with 40% lower costs · 13 src
- ChatGPT-6 Astra cracks 1941 Enigma message in two days, Crypto Cellar reports · 1 src
- Meta opens early access for new Muse AI features · 1 src
- GPT-6 Astra scores 99.9% on ARC-AGI-3, OpenAI says · 6 src
- Intern-Decision-4B released as multimodal structured decision model · 1 src
Comments
via GitHub Discussions