AI Daily · Sep 23, 2026: OpenAI releases GPT-6 Sol and Luna with half-price API access
OpenAI launches GPT-6 Sol and Luna with API prices halved, Luna at 1/100th cost of GPT-5.6. Anthropic releases Claude Opus 5.5 with 40% lower cost and raises usage limits.
Showing the latest briefing (Sep 23, 2026). Every day is generated the next morning.
Pick a date
-
OpenAI releases GPT-6 Sol and Luna with half-price API access
OpenAI launched GPT-6 Sol ($2/M input, $10/M output) and Luna ($0.10/M input, $0.50/M output) on OpenRouter, halving API prices; Luna scores 66.6% on DeepSWE v1.1 and matches GPT-5.6 Sol factuality at 1/100th cost.
-
Anthropic launches Claude Opus 5.5 with 40% lower running cost
Anthropic introduced Claude Opus 5.5, the first model in the Claude 5.5 family, matching Claude Fable 5.1 performance on most tasks while costing 40% less to run than Opus 5 at default settings.
-
Claude Opus 5.5 available via Requesty and OpenRouter with lower prices
Claude Opus 5.5 is now live on Requesty and OpenRouter. Requesty lists $4/M input and $20/M output (down from $5/$25) with cache reads at $0.20/M; OpenRouter offers 1M context at $4/M input and $20/M output, 20% lower per token than Opus 5.
-
Claude Opus 5.5 launches with Standard effort in Perplexity and Cursor support
Claude Opus 5.5 is now the Standard effort model in Perplexity Computer, scoring 0.610 on WANDR at $4.13 per task, 67.6% cheaper than Claude Fable 5.1. It also tops CursorBench at 57.8% (Max) with 40% lower cost per task than Opus 5, and LiteLLM reports it is 30% faster and 40% cheaper than Opus 5.
-
Anthropic raises five-hour usage limits and adds rate limit reset for subscribers
Anthropic is increasing five-hour usage limits on Pro, Max, and Team plans and giving subscription users a rate limit reset they can save and use whenever they choose.
-
GPT-6 Sol and Luna now available on B.AI with 1.05M context and DeepSWE scores
OpenAI's GPT-6 Sol and Luna are now live on B.AI's platform, featuring 1.05M context, 128K max output, vision, Computer Use, and six reasoning levels. Sol scores 68.8% on DeepSWE v1.1 with ~80% lower single-task cost; Luna scores 66.6%.
-
GPT-6 Sol and Luna roll out across ChatGPT, Codex, and API
GPT-6 Sol and Luna are available today in ChatGPT Work and Codex for Plus, Pro, Business, Enterprise, and Edu users, as well as in the API; free and Go users can try Luna in the desktop app.
-
Tencent Hy Image3.5 preview released with +30% win rate and 2K resolution
Tencent released Hy Image3.5 preview, reporting a +30% win rate in human evaluation vs Hy Image3.0, supporting text-to-image and image-to-image up to 2K resolution, with improved instruction following and consistency.
-
Qwen-Audio-3.1 launches with ASR-Next and TTS-Next, plus major price cuts
Qwen-Audio-3.1 upgrades its audio stack with five models, including new ASR-Next and TTS-Next, and cuts prices by up to 95% on ASR, ~70% on TTS, and ~85% on Realtime.
-
Xiaomi MiMo-V2.6 Series Available on B.AI with 1M Context
Xiaomi's MiMo-V2.6 open multimodal models, including Pro (1.02T total/42B active) and Flash (309B total/15B active), are now live on B.AI with API and web chat, featuring 1M token context and text/image/video/audio understanding.