New model releases, benchmarks, evaluation methods, and research results.
Google announced Gemini 4 Argon on September 30. Max output jumps from 64K to 1M tokens and 13 of the 19 rows in its own comparison table are outright wins, but there is no model ID, no launch date, and no public API at $2 per million input tokens.
Fireworks released Ember-1, a post-trained Kimi K3, on September 23 and shipped it to OpenRouter and Vercel AI Gateway on September 27. Per-token pricing matches K3 exactly at $3 in and $15 out, while token savings across five benchmarks range from 5.9% to 51.9%.
Anthropic released Claude Sonnet 5.5 on September 28. Pricing holds at Sonnet 5 levels, $2 per million input tokens, and Terminal-Bench 4.0 climbs from 10.3% to 70.6%, past Opus 5.5 at 66.4%.
OpenAI shipped GPT-6 Sol and Luna on September 22. Input runs $2 per million tokens on Sol and $0.10 on Luna, exactly half the predecessor rate, while the top DeepSWE score drops from 72.7% to 68.8%.
Anthropic released Claude Opus 5.5 on September 22. Input runs $4 per million tokens against Fable 5.1 at $10, and Terminal-Bench 4.0 climbs from 52.3% on Opus 5 to 66.4%.
PrismML released Ternary Bonsai 2 27B under Apache 2.0 on September 17, compressing Qwen3.8 27B into 5.9GB. The 20-benchmark average holds at 98.2% of the original, but long-horizon coding scores drop to three quarters.
Google shipped two speech-to-speech models on September 15. The cheap gemini-3.8-live lifts the composite index from 71.5% to 76.0%, but its tool-calling task completion rate lands below the model it replaces.
TypeSafe shipped Jev, a model that returns choices and probabilities instead of prose. Input runs $0.042 per million tokens with output free, but its own eval puts accuracy in the 67% range, roughly 6 points below GPT-5.6 Sol, and access is waitlist-only.
DeepSeek released the 552B open-weight V4.1 Flash on September 10 and will route V4 Pro calls to it from 04:00 UTC on September 14. Agentic coding beats Pro and output is 70% cheaper, but factual recall drops from 55.2 to 42.3.
ARC Prize ran GPT-6 Astra on ARC-AGI-3 under two harnesses and got 62.7% and 99.9%. Same model, same games. The only difference was whether reasoning state survived between calls. The leaderboard will now carry both numbers.
Nvidia Nemotron-3-Ultra-CC scored 535.4 at IOI 2026, above the 498.27 of the best human contestant. The paper breaks the score down, and most of it came from a test-time loop that resamples 200 candidates rather than from post-training.
MBZUAI-backed IFM released six K2 Horizon models under Apache 2.0 on September 3. The 7B beats Qwen3.5-9B on SWE-bench Verified at 70.6%, but the training-code repos and dataset links are still closed.