The small-model price war accelerated: Anthropic released Claude Haiku 5.5, its cheapest and fastest small model ever — roughly 75% cheaper to run than Haiku 4.5 ($0.10/M input, $0.50/M output under 100K tokens), with 1M context, adjustable effort, and improved alignment scores. Anthropic also halved Sonnet 5.5 cache-read pricing to $0.10/M tokens, making long-running agent workloads ~20% cheaper.
OpenAI rolled out GPT-6 with Intelligent UI across ChatGPT, letting the model answer with fully interactive interfaces — charts, buttons, calculators, and mini-apps generated inside the conversation rather than as plain text. Plus/Business tiers got GPT-6 Sol on October 7, with Free and Go tiers on GPT-6 Luna starting October 8. Google, meanwhile, opened SynthID Detector globally, a provenance portal that checks whether images, video, or audio were AI-generated — covering watermarks from Google, OpenAI, NVIDIA, and Kakao, with Apple to follow.
On the open-weights front, Mistral launched Mistral Large 4 ("Le Chonk"), a 1.05T-parameter multimodal MoE with 49B active parameters and 1M context, in public preview; open weights are promised by end of October. Arena's first real-world tests place it among the top-15 labs on code evaluations — the only European lab on that list.
OpenAI began rolling out GPT-6 across ChatGPT on October 7 with Intelligent UI, generating interactive graphics, forms, and tappable mini-apps inside conversations. Plus, Pro, Business, and Enterprise subscribers on GPT-6 Sol got it first; Free and Go tiers follow on GPT-6 Luna.
Haiku 5.5 targets high-volume subagent and browser workloads, cutting prices ~90% for requests under 100K tokens and matching GPT-6 Luna pricing. It is Anthropic's first Haiku with an adjustable effort setting, and shows fewer misaligned behaviors than Haiku 4.5.
Mistral's 1.05T-parameter open-weight multimodal MoE (49B active, 1M context) is available now via preview API, with open weights promised by end of October. Early Arena testing places Mistral among the top 15 labs on code — the only European lab on the list.
The portal at synthid.com now checks images, video, and audio for AI watermarks from Google and partners including OpenAI, NVIDIA, and Kakao (Apple to follow). It is available globally in English as of October 7, after earlier testing with journalists and media professionals.
Two late-interaction embedding models (9B and 0.6B) retrieve text, images, and pages in a shared embedding space, enabling cross-model querying: a corpus indexed with the 9B model can be searched using 0.6B queries. Both are publicly available on Hugging Face.
Selected posts from the edition, with links to the originals.
"Introducing Claude Haiku 5.5: the cheapest, fastest, and most capable small model we've ever released. On average, it costs around 75% less to run than Claude Haiku 4.5."
View post on X ↗"ChatGPT can now generate a custom UI for you!"
View post on X ↗"RT @GoogleDeepMind: SynthID Detector is now available to everyone. 🌐 Check whether online content was generated using @GoogleAI, or with t…"
View post on X ↗"Le Chonk is here, and so are first impressions from @petergostev. Mistral Large 4 by @MistralAI is a 1T-parameter model (49B active), with open weights coming at the end of the month. We put it through a series of real-world tests…"
View post on X ↗"We're releasing pplx-embed-v2-late, two late-interaction embedding models that retrieve text, images, and pages with a shared embedding space for cross-model querying. Both models achieve frontier performance and are publicly available on Hugging Face."
View post on X ↗"Claude Haiku 5.5 is now available in Devin. On FrontierCode 1.1, Haiku 5.5 scores 58.4%. That puts it ahead of Sonnet 5 at about an eighth of the cost per task."
View post on X ↗"RT @steipete: I hooked up our team claw to X to trigger work faster. Unassigned sessions are for anyone to grab. Our agent looks who worked…"
View post on X ↗"Built alongside the blind and low-vision community, Guided Vision in Gemini Live offers conversational, real-time visual assistance. Now you can share your camera to receive dyna…"
View post on X ↗"We evaluated Jev Router by @typesafeai on Agent Arena. Our tests spanning more than 4,700 real-world agentic sessions show: 1. Jev Router does not improve on the current Pareto frontier. For similar performance as DeepSeek V4.1 Flash (Max), the solution costs 38% more…"
View post on X ↗"RT @SemiAnalysis_: A $200 AI subscription can be worth $12,000 in tokens. Which plan delivers it is not close. 'They're selling to willing…'"
View post on X ↗