AI digest of the week: Claude Haiku 5.5, Mistral Large 4, Gemini and Odyssey-3

AI digest of the week

LLMs

• Claude Haiku 5.5 — $0.10/$0.50 per million input/output tokens for prompts up to 100K. On longer requests, both prices increase fivefold.
• Mistral Large 4 — 1 trillion parameters, 52B active. Currently an API preview; weights are promised by the end of October.
• Ultrafast for Sol — an accelerated GPT-6.1 Sol has appeared in Codex/Work on Pro for $500. In the API on short context, it costs $12/$60 per million input/output tokens.
• Math from OpenAI — hundreds of preprints from an unreleased model after attempts to solve about 4,000 problems. Some proofs have been formalized in Lean, while others may still contain errors.
• OpenAI watermarks — in the API, text marking can be enabled voluntarily; in ChatGPT and Codex in the EU, it will be added in the coming weeks. Short texts and math are poorly detectable.
• Free Gemini — starting October 9, Google is removing Flash and Pro from free accounts. Only Flash-Lite will remain. AI Plus will keep Flash; AI Pro and Ultra will keep all three models.

Generative models

• Odyssey-3 — generates a world from a prompt and reacts to user actions in real time. You can walk around as a cat, kayak, and change events on the fly; Flash is available in the browser.
• Nano Banana 2.1 — an update based on Gemini 3.6 Flash. Outputting a 1K image costs $0.0336 without input and thinking, but in the Edit Drift test the image still drifts after a series of edits.