Kolibri-1 serves 1M tokens but was trained to 256K
Aleph Alpha released Kolibri-1 on 3 October 2026 under Apache 2.0. The headline says 1M tokens of context. The model card says it was trained on sequences up to 256K, ...
Tag: news
Gemini 4 Argon costs $2 and $10 now, $4 and $20 later
Google listed Gemini 4 Argon at $2 input and $10 output per million tokens on 30 September 2026, with a later rate of $4 and $20. Almost nobody can call ...
GPT-6.1 Sol costs what GPT-6 Sol did, and only the cache got cheaper
A fifth of Astra’s price, yes. Against GPT-6 Sol, the only price that moved is cached input, cut from $0.20 to $0.10.
Claude Sonnet 5.5 keeps $2 and $10 and retires thinking disabled
Same $2 and $10 price as Sonnet 5, benchmarks within two or three points of Opus 5.5, and a disabled thinking setting that now returns a 400.
MiMo-V2.6-Pro was quietly retrained to stop looping on tool calls
Same model names, new weights. Xiaomi retrained MiMo-V2.6-Pro and Flash to stop repeating tool calls, shipped it to the API first and the MIT weights two days later. We read ...
Anthropic’s $11.6B Akamai deal buys CPUs, not GPUs
Seven years, $11.6 billion, up to about $20 billion with options, and not one GPU. We read Akamai's 8-K and the call coverage to see what Anthropic is actually buying.
Google’s first Suncatcher satellite flies four TPUs on 1 kW
Google’s first Project Suncatcher satellite is a survival test: four Trillium TPUs, about a kilowatt of solar power, bursts of roughly 15 minutes before cooling. The laser links wait for ...
Claude Opus 5.5 drops to $4 and $20, and breaks four Opus 5 habits
Anthropic cut Opus to $4 and $20 per million tokens and cache reads by 60%. The catch is four request shapes that Opus 5 accepted and Opus 5.5 rejects.
Grok 4.7 keeps $2 and $6, and its gains over 4.6 are xhigh versus high
xAI shipped Grok 4.7 on 21 September at the same $2 and $6 per million as 4.6, same 500K window, same 200K pricing band. Its own table shows gains on ...
Qwen-Image-2.1 brings the weights back, but not the Apache licence
The Qwen image line has downloadable weights again: a 7B transformer, an 8B text encoder and an RGBA VAE. The licence is research or evaluation only, a first for the ...