DeepSeek V4-Flash Gets Vision — Experimental Image Input Comes to the Budget Model
DeepSeek quietly ships deepseek-v4-flash-vision-exp, an experimental vision variant of its V4-Flash model that accepts image input, alongside V4-Pro and V4-Flash point updates.
GPT-6 Astra: OpenAI Claims New Intelligence Generation — 98% FrontierMath, Human Parity on ARC-AGI-3, and New Prime Number Proofs
OpenAI's GPT-6 Astra capabilities announcement: state-of-the-art on computer use, coding, and science with saturated benchmarks, two real prime-gap proofs, and 0% rate of escaping authorized scope vs Sol's 48%.
OpenAI Releases GPT-6 Astra — Its Most Capable Model Ever, and the First at Critical Cybersecurity Level
OpenAI launches GPT-6 Astra, the most capable model it has broadly deployed, with major alignment improvements, jailbreak robustness, and a candid warning that monitorability has decreased.
Introducing Claude Fable 5.1 and Claude Mythos 5.1 — Anthropic's Most Advanced Models for Coding and Knowledge Work
Anthropic releases Claude Fable 5.1 and Mythos 5.1, its most advanced models for coding and knowledge work, with research capabilities offering an early glimpse of AI contributing to scientific progress.
Google Ships Gemini 3.7 Flash — Its Smartest Workhorse Yet
Google releases Gemini 3.7 Flash, its most capable workhorse model, arriving just three weeks after 3.6 with improved reasoning, faster inference, and better tool use.
Grok 4.6: Built for Long-Running Agents and Visual Work
xAI's Grok 4.6 launches with a focus on long-running agents and more ambitious interactive and visual work, introduced first through Cursor's platform.
Google Delays Gemini 3.5 Pro Launch to Bolster Programming Capabilities
Google postponed Gemini 3.5 Pro after internal tests fell short of goals in code generation, as OpenAI and Anthropic accelerate — Alphabet shares dipped nearly 3%.
OpenAI Launches GPT-5.6: Sol, Terra, and Luna Model Family Arrives After US Government Review
OpenAI has released GPT-5.6, its new model family of Sol, Terra, and Luna tiers, after a weeks-long US government cybersecurity review cleared the models for public rollout.
Meta Releases Muse Spark 1.1: A Purpose-Built Model for Agentic Tasks and Tool Use
Meta launches Muse Spark 1.1, an AI model designed specifically for agentic tasks like tool usage, code execution, and multi-step workflows. The model targets enterprise automation and competes directly with Claude and GPT-5.6.
xAI Releases Grok 4.5: Opus-Class Performance at Half the Price of Competitors
Grok 4.5 targets coding, agentic tasks, and knowledge work at $2 per million input and $6 per million output tokens — roughly half the price of leading rivals.
Anthropic Launches Claude Sonnet 5: First Sonnet-Class Model to Outscore Opus 4.8
Anthropic released Claude Sonnet 5 on June 30, 2026, making it the default model for Free and Pro plans. Priced at $2 per million input tokens, it is the first Sonnet-class model to outscore the Opus 4.8 flagship.
NVIDIA Accelerates Google DeepMind’s DiffusionGemma for Local AI
Google DeepMind released DiffusionGemma, an experimental open model that denoises up to 256 tokens at once, and NVIDIA optimized it to run up to 4x faster on RTX and DGX systems.
The Latest AI News We Announced in May 2026
Google's May 2026 AI recap covers the agentic Gemini era - the Gemini 3.5 and Gemini Omni launches plus the Googlebook, Fitbit Air, Universal Cart, and a new quantum-AI life sciences program.
9 demos of Gemini Omni and Gemini 3.5 in action
Nine videos from Google I/O 2026 show Gemini Omni editing video through conversation and Gemini 3.5 Flash powering agentic coding, generative UI, information agents, and Gemini Spark.
Gemini 3.5: Frontier Intelligence with Action
At Google I/O 2026, Google launches Gemini 3.5 Flash, an agentic coding model that beats Gemini 3.1 Pro on key benchmarks and runs 4X faster than frontier rivals.
Introducing the Ettin Reranker Family
Hugging Face releases six CrossEncoder rerankers built on the Ettin ModernBERT encoders, from 17M to 1B parameters — state-of-the-art at their sizes on MTEB retrieval and up to 8.3x faster than prior baselines.
OpenAI Launches GPT-5.5 as Most Advanced Model Yet
OpenAI announces GPT-5.5, positioning it as the most advanced model for real-world work and agentic AI.
OpenAI Launches GPT-5 With Breakthrough Reasoning
OpenAI's latest model GPT-5 introduces advanced chain-of-thought reasoning, outperforming humans on complex benchmarks.