Mistral Large 4 and Claude Haiku 5.5 headline new AI model releases

Mistral shipped a public preview of Large 4, a natively multimodal 1-trillion-parameter MoE model (52B active) trained on its own 3,800-GPU Grace Blackwell cluster. Anthropic launched Claude Haiku 5.5, roughly 75% cheaper than Haiku 4.5 but with a tokenizer that uses about 1.25x more tokens. OpenAI added an Ultrafast tier for GPT-6.1 Sol and a beta Decisions API.

DEV Community

Open source

Mistral Large 4 and Claude Haiku 5.5 headline new AI model releases

Mistral Large 4: 1T parameters, 52B active, 160+ languages, open weights promised by end of month
Claude Haiku 5.5 is ~75% cheaper than Haiku 4.5 but its tokenizer uses ~1.25x more tokens
OpenAI's Decisions API on gpt-6-luna claims ~10x faster responses than the Responses API
LightOnOCR-3 (0.8B-4B, Apache 2.0) scores 86.3 on olmOCR-Bench, ~19% faster than Chandra-OCR-2

#Mistral#Anthropic#OpenAI#Claude

Mistral Large 4 and Claude Haiku 5.5 headline new AI model releases

AI Dev Weekly #29: Haiku 5.5, Mistral Large 4, Decisions API and Copilot

Mistral releases Large 4, a 1T-parameter open-weight AI model

Claude Haiku 5.5 Consumes 3X More Tokens Than ChatGPT 6 Luna

Anthropic launches Claude Haiku 5.5 with up to 90% price cut