370 summarised stories about Model Release, each linking back to the original source. Browse all topics →
Latent Space 1 week ago 3 ● 4 sources
OpenAI fired three AI safety researchers tied to the METR/Hugging Face incident and they disputed the reasons while OpenAI reportedly said they mishandled confidential information. Pricing for OpenAI’s GPT-6.1 Sol sets input/output tokens at $12 and $60 per 1M tokens. The updates include broader model rollouts and multiple safety and evaluation releases, plus ongoing debate about agent monitorability and authorization norms.
Trending Topics 1 week ago 35 ● 16 sources
Anthropic launched Claude Haiku 5.5 and cut pricing for its small model. The API price drops to $0.10 per million input tokens (and $0.50 per million output tokens), but only for requests up to 100,000 tokens, and a new tokenizer can count text as about 30% more tokens. Pricing for typical workloads is expected to be about 75% lower, alongside added capabilities like adjustable effort and updated safety and availability.
Anthropic 1 week ago 5 ● 16 sources
Anthropic released Claude Haiku 5.5, a smaller Claude model aimed at high-volume, speed-sensitive tasks and subagent coding work. The model costs around 75% less to run than Haiku 4.5, and Anthropic also cut Sonnet 5.5 cache read pricing by 50% to make it about 20% cheaper on most agentic work. Anthropic is also adding an adjustable “effort” setting and introducing new monthly API credits for Max and Team subscribers, with platform SDK updates adding beta support for computer and browser use.
Shtetl-Optimized 1 week ago 3 ● 19 sources
OpenAI released 372 major mathematical results, including a proof of Subhash Khot’s Unique Games Conjecture, but the author says few humans yet understand the proofs and are starting a new digestion effort. The work reportedly used about 3 hours of GPT-Pro-level compute on average per solved problem and tested on about 8,000 problems. As a result, mathematicians are rushing to formalize and learn from messy AI-generated proofs while the field shifts toward using AI tooling to help verify and explain progress.
TLDR 1 week ago 43 ● 16 sources
Anthropic released Claude Haiku 5.5, a fast, low-cost model. It charges $0.10 per million input tokens and $0.50 per million output tokens up to 100,000 tokens. Pricing and value shift by length, with Haiku reported to be a better deal under 100,000 tokens and a worse deal beyond that.
VentureBeat 1 week ago 19 ● 16 sources
Anthropic launched Claude Haiku 5.5 and lowered API prices for the model. The article says it matches OpenAI’s GPT-6 Luna to the penny. Average job costs dropped versus the prior Haiku version, and cache-read pricing was cut in half.
The Batch 45
Open-weight model GLM-5.3 was analyzed as approaching Claude Mythos’ cyber capability, including a 12% versus 14% success rate on ExploitBench tasks at comparable token counts and evidence that $20.40 of GLM-5.3-Flash tokens could find a recently disclosed Chrome flaw. Xiaomi then released MiMo-V2.6-Pro-RL, MiMo-V2.6-Flash, and a distilled variant, with MiMo-V2.6-Pro-RL costing $0.435 per million input tokens (and free downloads under MIT) while reporting RL training details and CyberBench v1.1 results like 75.36% accuracy for MiMo-V2.6-Flash. Defenders are pushed to accelerate sandboxing, monitoring, and vulnerability mitigation as attackers can access strong open-weight cyber-capable models and the cost to generate attacks is dropping.
Latent Space 1 week ago 22 ● 16 sources
Anthropic released Claude Haiku 5.5, updating its Haiku-tier small model after about a year. The new model is priced to match OpenAI’s GPT-6 Luna at the same rates and Anthropic also cut Sonnet 5.5 cache-read pricing from $0.20 to $0.10 per 1M tokens. As a result, Anthropic expands Haiku’s role as the cheaper, high-volume subagent for multi-model agent workflows while also lowering related operating costs and adding effort settings plus a 1M-token context.
The daily briefing
TLDRocket reads all relevant sources, removes duplicate coverage, and summarises the day in two minutes. Follow companies and topics for alerts, or get the briefing in Slack. Free, no spam, unsubscribe anytime.
TLDRocket
17 October 2026
AI governance moved from slideware to the boring-but-important mechanics of who gets to fetch, act, and prove. Meta quietly rewrote parts of its commercial terms, removing wording that appeared to let it use web content to “train” and “maintain the integrity” of its products; the update followed legal scrutiny that it could be read as broad scraping permission. At the same time, users reported that major websites are blocking AI agents from completing tasks across services—turning “agentic” demos into something closer to permissioned automation. ServiceNow is now trying to sell the glue for that reality with its AI Control Tower, positioning it as an enterprise platform to discover and govern agents across models and clouds, while a field study of 15 large-enterprise customers questioned how fully its runtime enforcement and AI lifecycle evidence hold up versus alternatives.
Read the full briefing →