Claude Haiku 5.5 (Anthropic)
Claude Haiku 5.5 at 10 cents per million input tokens
AI Year in Review · Oct 10, 2025 – Oct 10, 2026
Models & LLM releases
122 stories in this category.
GLiDE (Fastino)
Fastino GLiDE, a thinking decision model
Mistral Large 4 (Mistral AI)
Mistral Large 4 "Le Chonk": 1T parameters, open weights promised for October
Beam (Reflection AI)
Reflection AI announces Beam, a 501B Western open-weight model
Rho-1 (Reka)
Reka Rho-1, a 19B research-preview omni model
GPT-6 in ChatGPT for everyone (OpenAI)
GPT-6 Sol (paid) & GPT-6 Luna (Free/Go) roll out to all ChatGPT users with "Intelligent UI".
Claude Sonnet 5.5 (Anthropic)
Anthropic ships Claude Sonnet 5.5, beating Opus 5.5 on Terminal-Bench 4.0
Gemini 4 Argon (Google DeepMind)
Google's Gemini 4 Argon tops Text Arena and the Vals Index, trusted testers only
D1 (Liquid AI)
Liquid AI releases D1, its first decision model
GPT-6.1 Sol (OpenAI)
OpenAI releases GPT-6.1 Sol: near-Astra intelligence, cached input 95% off
Perceptron Mk1.5 (Perceptron)
Perceptron Mk1.5 arrives on OpenRouter
Span-01 (Respan)
Respan's Span-01 decision model claims half the price of Jev
Gemini 4 Argon
Google released Gemini 4 Argon first to trusted cyber defenders, with a 1-million-token output limit . The same day, Anthropic’s release notes deprecated Claude Sonnet 4.5, with retirement on November 30, 2026.
GPT-6.1 Sol
OpenAI upgraded Sol a week after its release.
OpenAI DevDay
OpenAI’s DevDay recap covers GPT-6.1 Sol, always-on Dots agents and an Agents API with Computer Use .
Claude Sonnet 5.5
Anthropic released Sonnet 5.5, the first Sonnet model to launch with cyber safeguards .
Qwen 3.8 Omni Flash (Alibaba (Qwen))
Qwen 3.8 Omni Flash with 1M context and Qwen 3.8 Live Translate
Monthly roundup: ThursdAI Sep 2026 ↗
Monthly roundup
Claude Opus 5.5 (Anthropic)
Anthropic launches Claude Opus 5.5: Fable-level work for 40% less
Monthly roundup: ThursdAI Sep 2026 ↗
Monthly roundup
GPT-6 Luna (OpenAI)
OpenAI releases GPT-6 Luna at $0.10 / $0.50 per million tokens
Monthly roundup: ThursdAI Sep 2026 ↗
Monthly roundup
GPT-6 Sol (OpenAI)
OpenAI rolls out GPT-6 Sol at half the GPT-5.6 price
Monthly roundup: ThursdAI Sep 2026 ↗
Monthly roundup
Grok 4.7 (xAI)
xAI releases Grok 4.7 with 46% on CursorBench
Monthly roundup: ThursdAI Sep 2026 ↗
Monthly roundup
Meta Connect
Meta said its Muse agent is coming to its AI glasses “in the coming months.”
Claude Opus 5.5
Anthropic released Opus 5.5 with a 1-million-token context window . Artificial Analysis scored it 58 on its Intelligence Index at max effort.
GPT-6 Sol and GPT-6 Luna
OpenAI added two GPT-6 models in ChatGPT Work, Codex and the API .
xAI Grok 4.7
xAI released Grok 4.7 with a 500,000-token context window and reported 71.0% on DeepSWE v1.1 at high effort in its own testing.
Union Alpha (OpenRouter)
Union Alpha: anonymous stealth model free on OpenRouter with 262K context
Jev (TypeSafe AI)
TypeSafe AI launches Jev, the first public non-LLM System 1 decision model
SWE-2 (Cognition)
Cognition SWE-2: near-frontier coding at up to 70% lower cost, free for a month on Devin
DeepSeek-V4.1-Flash
DeepSeek’s change log calls it “the smallest model in our new architecture family, with native multimodal visual understanding”; it replaces V4 Flash and V4 Flash Vision Exp.
Qwen3.8-Max-0902 (Alibaba Qwen)
Qwen3.8-Max-0902: 2.4T API-only refresh claims #1 on Code Arena
Claude Fable 5.1 & Mythos 5.1 (Anthropic)
Claude Fable 5.1 & Mythos 5.1: Terminal-Bench 4.0 jumps to 55.8, cache reads 75% cheaper
Gemini 3.8 Flash (Google DeepMind)
Gemini 3.8 Flash: third Flash in three weeks, HLE-Verified 54.9, 1M context
Muse Spark 1.3 (Meta AI)
Meta Muse Spark 1.3 ties GPT-5.6 Sol and Grok 4.6 on the AA index, max mode ties Fable 5
GPT-6 Astra (OpenAI)
OpenAI launches GPT-6 Astra, calls it its most intelligent and aligned model
Solaris (Runway)
Runway Solaris: an Interface World Model that generates clickable UIs frame by frame
Gemini 3.8 Flash and 3.8 Flash Cyber
Google released the pair: Flash across its developer tools and consumer products, Flash Cyber only for trusted defenders through its Fairwind Program.
Meta Muse Spark 1.3
Meta announced Muse Spark 1.3, available on Muse Code and the Meta Model API.
Claude Fable 5.1 and Claude Mythos 5.1
Anthropic released both, saying they “are the same model, but with different levels of safeguards.” Fable 5.1 is generally available; Mythos 5.1 is limited to vetted organizations.
Navigator n2 (Yutori)
Yutori Navigator n2: 27B computer-use model at a fraction of frontier cost
OpenAI speech-to-text deprecations
OpenAI’s deprecations page lists whisper-1, gpt-4o-transcribe, gpt-4o-mini-transcribe and gpt-4o-transcribe-diarize as deprecated, with shutdown on February 26, 2027. The recommended replacements are gpt-li
DeepSeek-V4-Flash-Vision-Exp
The same change log adds DeepSeek-V4-Flash-Vision-Exp as an experimental model in the API.
Gemini 3.7 Flash (Google DeepMind)
Gemini 3.7 Flash: 300+ tok/s mid-tier multimodal with a 50% price cut
Grok 4.6 (xAI)
Grok 4.6 ties GPT 5.6 Sol at half the price
Gemini 3.7 Flash
Google’s Gemini API changelog lists Gemini 3.7 Flash as generally available in the Gemini API, with a 1,048,576-token context window and text, image, video, audio and PDF input.
DeepSeek-V4-Pro
DeepSeek’s API change log records DeepSeek-V4-Pro as generally available across the DeepSeek app, web and API.
Claude Opus 4.1 retired
Anthropic’s release notes list Claude Opus 4.1 as retired: requests to the model now return an error, and Anthropic recommends Claude Opus 5 as the replacement.
Qwen3.8-Max (Alibaba (Qwen))
Qwen3.8-Max: Alibaba's 2.4T-parameter flagship, with open weights promised within a week
DeepSeek-V4-Flash
DeepSeek’s API change log records DeepSeek-V4-Flash as a public beta in the DeepSeek API.
Claude Opus 5 (Anthropic)
Anthropic ships Claude Opus 5 at unchanged Opus pricing, claiming near-Fable-5 performance at half the cost
Ling-3.0-Flash (Ant Group)
Ant Group's Ling-3.0-Flash: a 124B MoE that claims to match its 1T flagship on ~5B active parameters
Gemini 3.6 Flash / 3.5 Flash-Lite / 3.5 Flash Cyber (Google DeepMind)
Google ships a three-model Gemini Flash refresh — but still no Gemini 3.5 Pro
Gemini 3.6 Flash and Gemini 3.5 Flash-Lite
Google’s Gemini API changelog lists both models as generally available in the Gemini API, each with a 1,048,576-token context window and text, image, video, audio and PDF input.
Qwen3.8-Max Preview (Alibaba Qwen)
Alibaba previews Qwen3.8-Max at a claimed 2.4T parameters, 'second only to Fable 5'
Moonshot AI Kimi K3
Moonshot introduced Kimi K3, a 2.8-trillion-parameter mixture-of-experts model with a 1-million-token context window and native image and video input. The weights are published on Hugging Face under the Kimi K3 License.
Muse Spark 1.1 & Meta Model API (Meta AI)
Meta launches Muse Spark 1.1 and its first paid Meta Model API
GPT-5.6 (Sol, Terra, Luna) (OpenAI)
OpenAI launches GPT-5.6 publicly as three tiers: Sol, Terra and Luna
OpenAI GPT-5.6
OpenAI released GPT-5.6 in three tiers: Sol, which it calls “our flagship,” Terra, “a balanced model for everyday work,” and Luna, which the API changelog positions “for efficient, high-volume workloads.” All three are in the Ope
SWE-1.7 (Cognition)
Cognition ships SWE-1.7 at 1000 tokens per second
Grok 4.5 (xAI)
SpaceXAI launches Grok 4.5, a coding-and-agents model trained with Cursor
Base 1 (Base44)
Base44 launches Base 1, the first in-house vibe-coding LLM
Monthly roundup: ThursdAI Jul 2026 ↗
Monthly roundup
GPT-5.6 (OpenAI)
OpenAI ships GPT-5.6 as a three-model family: Sol, Terra and Luna
Monthly roundup: ThursdAI Jul 2026 ↗
Monthly roundup
Fable 5 (Anthropic)
Fable 5 restored globally after the export-control pause
Monthly roundup: ThursdAI Jul 2026 ↗
Monthly roundup
Sonnet 5 (Anthropic)
Claude Sonnet 5: 'our most agentic Sonnet yet' at intro pricing
Claude Sonnet 5 (Anthropic)
Most agentic Sonnet, near Opus 4.8 at $2/$10; default for Free & Pro.
Claude Fable 5 & Mythos 5 (Anthropic)
Anthropic’s most capable GA model ($10/$50), plus Mythos tier above Opus with 1M context.
MAI-Code-1-Flash (Microsoft)
Microsoft ships MAI-Code-1-Flash into GitHub Copilot
MAI-Thinking-1 (Microsoft)
Microsoft launches MAI-Thinking-1, a 1T MoE trained from scratch
MiniMax M3 (MiniMax)
MiniMax announces M3 coding/agentic model with 1M context
Claude Opus 4.8 (Anthropic)
Anthropic ships Claude Opus 4.8 live mid-show
Qwen 3.7-Max (Alibaba (Qwen))
Alibaba releases Qwen 3.7-Max agentic frontier model with robotics demos
Composer 2.5 (Cursor)
Cursor launches Composer 2.5 with Opus-class coding at much lower cost
Gemini 3.5 Flash (Google DeepMind)
Gemini 3.5 Flash launches at I/O as Google's agentic workhorse model
Google I/O 2026: Gemini 3.5 Flash, Gemini Omni, Spark
"Agentic Gemini era": Gemini 3.5, Omni (reason+create), Search information agents, Antigravity 2.0.
ERNIE 5.1 Preview (Baidu)
Baidu ERNIE 5.1 Preview hits #13 on Arena with 6% of the compute
REDMOD (Mayo Clinic)
Mayo Clinic's REDMOD detects pancreatic cancer 3 years early
OpenAI clinician model + workspace agents (OpenAI)
OpenAI releases clinician/medical model and workspace agents
Monthly roundup: ThursdAI Apr 2026 ↗
Monthly roundup
GPT-5.5 (OpenAI)
GPT-5.5 and GPT-5.5 Pro drop live, SOTA across the board
GPT-5.5 & GPT-5.5 Pro (OpenAI)
82.7% Terminal-Bench 2.0, 58.6% SWE-Bench Pro; API on Apr 24.
Claude Opus 4.7 (Anthropic)
Claude Opus 4.7 drops live with 87.6% SWE-bench Verified and xhigh effort
Muse Spark (Meta (Meta Superintelligence Labs))
Meta launches Muse Spark, first model from Meta Superintelligence Labs
Qwen3.6-Plus (Alibaba (Qwen))
Alibaba ships Qwen3.6-Plus with near-Opus agentic coding and 1M context
Composer 2 (Cursor)
Cursor Composer 2 beats Opus 4.6 on TerminalBench at a tenth of the price
MiniMax M2.7 (MiniMax)
MiniMax M2.7: first self-evolving model hits 56% on SWE-Bench Pro
GPT-5.4 Mini & Nano (OpenAI)
OpenAI ships GPT-5.4 Mini and Nano for coding, computer use, and subagents
MiMo (Xiaomi)
Xiaomi MiMo revealed as the 1T-param stealth model topping OpenRouter
GPT-5.4 mini & nano (OpenAI)
Cheaper, faster GPT-5.4 variants.
Gemini Embedding 2 (Google)
Google launches Gemini Embedding 2, a natively multimodal embedder
embed-large-v3 (Mixbread)
Mixbread embed-large-v3 beats Gemini Embedding 2
SWE-1.6 (Cognition)
Cognition previews SWE-1.6, hitting 51% on SWE Bench Pro
Gemini 3.1 Flash-Lite (Google DeepMind)
Google launches Gemini 3.1 Flash-Lite with 1M context at 360 tok/s
GPT-5.3 Instant (OpenAI)
OpenAI rolls out GPT-5.3 Instant as the free-tier fast model
GPT-5.4 (OpenAI)
OpenAI drops GPT-5.4 Thinking and GPT-5.4 Pro live during the show
Claude Sonnet 4.6 (Anthropic)
Anthropic ships Claude Sonnet 4.6 with 79.6% SWE-Bench and 1M context
Seed 2.0 (ByteDance)
ByteDance Seed 2.0: frontier multimodal family at 73-84% lower pricing
Gemini 3.1 Pro (Google DeepMind)
Gemini 3.1 Pro drops live with 44% HLE and 77% ARC-AGI at the same price
Grok 4.20 (xAI)
xAI silently drops Grok 4.20 with four 500B-param collaborating agents
Gemini 3.1 Pro (Google)
Reasoning upgrade, 77.1% ARC-AGI-2.
Claude Sonnet 4.6 (Anthropic)
Sonnet upgrade with 1M context (beta), coding & computer use.
Gemini 3 Deep Think (Google DeepMind)
Gemini 3 Deep Think scores 84% on ARC-AGI 2
GPT 5.3 Codex Spark (OpenAI)
OpenAI ships GPT 5.3 Codex Spark on Cerebras for real-time coding
Claude Opus 4.6 (Anthropic)
Anthropic ships Claude Opus 4.6 with 1M context and agent teams
GPT-5.3-Codex (OpenAI)
OpenAI answers Opus with GPT-5.3-Codex, first model that helped build itself
Claude Opus 4 (Anthropic)
Claude Opus 4 drops in Q2 — Ryan's pick for best model ever
Monthly roundup: ThursdAI Dec 2025 ↗
Monthly roundup
Gemini 2.5 (Google DeepMind)
Gemini 2.5 takes the #1 benchmark spot in March
Hailuo 2.3 (MiniMax (Hailuo))
MiniMax drops Hailuo 2.3 in November
Monthly roundup: ThursdAI Dec 2025 ↗
Monthly roundup
GPT-5 Codex (OpenAI)
GPT-5 Codex: OpenAI's specialized coding model moves the stock
Monthly roundup: ThursdAI Dec 2025 ↗
Monthly roundup
Gemini 3 Flash (Google DeepMind)
Gemini 3 Flash delivers frontier intelligence at $0.50/1M input tokens
Mistral OCR 3 (Mistral AI)
Mistral OCR 3 claims 74% win-rate over OCR v2 with aggressive pricing
GPT 5.2 Codex (OpenAI)
GPT 5.2 Codex drops live during the show with 400K context
Gemini 3 Flash (Google)
Fast/cheap tier with strong baseline.
GPT-5.2 (OpenAI)
Instant/Thinking/Pro after "code red"; 100% AIME 2025, 70.9% GDPval.
Claude Opus 4.5 (Anthropic)
Anthropic launches Claude Opus 4.5, reclaiming the coding crown
Gemini 3 Pro (Google DeepMind)
Gemini 3 Pro launches with record ARC-AGI-2 scores
Monthly roundup: ThursdAI Nov 2025 ↗
Monthly roundup
GPT-5.1-Codex-Max (OpenAI)
GPT-5.1-Codex-Max runs 24-hour coding tasks with native compaction
Monthly roundup: ThursdAI Nov 2025 ↗
Monthly roundup
ACT-1 & Memo (Sunday Robotics)
Sunday Robotics unveils ACT-1 home robot foundation model and Memo
Monthly roundup: ThursdAI Nov 2025 ↗
Monthly roundup
Grok 4.1 (xAI)
Grok 4.1 briefly tops LM Arena with major post-training upgrade
Monthly roundup: ThursdAI Nov 2025 ↗
Monthly roundup
GPT-5.1 (OpenAI)
OpenAI launches GPT-5.1 with a warmer, more personable voice
SWE-1.5 (Cognition)
Cognition SWE-1.5: 950 tok/s coding model hitting 40% on SWE-bench Pro
Claude Haiku 4.5 (Anthropic)
Claude Haiku 4.5: fast, cheap model rivals Sonnet 4 accuracy
SWE-grep (Cognition)
Cognition SWE-grep: RL-trained fast context retrieval for coding agents
OpenPipe Qwen3 14B Instruct (OpenPipe (Weights & Biases))
OpenPipe Qwen3 14B Instruct lands on W&B Inference
RTFM (World Labs)
World Labs RTFM renders 3D worlds in real time on a single H100