Model launchesNew tools and SubQ's sparse-attention model aim to scale long-context LLMs and agentsA set of recent releases and tutorials focus on making large language models and agents more efficient with long contexts.5 min read
Model launchesOpenAI makes the GPT-5.5 Instant model the default; personalization and memory features expandOpenAI will make GPT-5.5 Instant the default model for all ChatGPT users within the next two days and will make it available in the API as 'gpt-5.5-chat-latest'; additionally, personalization improvements are coming for Plus and Pro web users, and memory sources will be available on all consumer plans on the web, with mobile availability coming soon.1 min read
Model launchesOpenAI introduces the GPT-5.5 Instant model in ChatGPTOpenAI announced that the rollout of GPT-5.5 Instant in the ChatGPT service has begun; the new model provides smarter, clearer and more personalized responses in a warmer, more natural tone, while…1 min read
Model launchesChallenges to Dario Amodei’s Safety Narrative from Jensen Huang and GPT-5.5 ResultsAnthropic CEO Dario Amodei’s warnings about AI-driven job losses and the need to withhold models from the public have been questioned from two directions: NVIDIA CEO Jensen Huang publicly pushed back against apocalyptic labour claims, and results released by AISI show the public GPT-5.5 model performing close to a restricted model called Mythos on expert cyber tasks.3 min read
Model launchesAnthropic's Mythos model shown to select US leaders; Portfolio Checklist discusses AI risks and agricultural damage from spring weatherThe latest Portfolio Checklist podcast covers Anthropic’s new AI model, Mythos, which the company has revealed only to a very small group of US stakeholders including Pentagon officials and senior bank executives due to its unexpectedly powerful capabilities.2 min read
Model launchesxAI’s Grok 4.3 Improves but Still Trails Leading ModelsxAI released Grok 4.3 with lower API prices, faster response, better tool integration and higher benchmark scores than prior Grok versions.2 min read
Model launchesOpenAI: strongest model launch in a week — GPT-5.5 accelerated revenue growthOpenAI announced that GPT-5.5 represents its strongest model launch to date one week after release; API revenue is growing more than twice as fast as with previous releases, and Codex revenue doubled in less than seven days.1 min read
Model launchesAlibaba’s HappyHorse Scores Moderately After Gray TestingAlibaba’s HappyHorse video model entered gray testing on April 27 after leading Artificial Analysis’s blind video arena over Seedance.2 min read
Model launchesOpenAI’s GPT-5.5 Boosts Benchmark Scores but Shows High Rate of Confident ErrorsOpenAI’s GPT-5.5 advances to state‑of‑the‑art results on several objective benchmarks—especially in knowledge, agentic tasks and abstract visual reasoning—while simultaneously producing a high proportion of incorrect but confident answers.4 min read
Model launchesMoonshot AI releases Kimi K2.6 — a 1T-parameter open vision-language model for long-running autonomous codingMoonshot AI has released Kimi K2.6, a 1 trillion-parameter vision-language model that supports very long autonomous code generation loops and larger multi-agent orchestration.5 min read