Friday, October 9.
Everything that shipped in AI.
27 stories, 63 lab updates, 36 things people built, 337 models ranked. Every one links to where it came from.
Tracking
Top story · Mistral AI · Oct 6
Mistral Large 4 is out, and the weights are open
- parameters
- 1.05T
- active per token
- 52B
- token context
- 1M
The brief · 26 more
Today, line by line.
- 01Claude Code Projects opens to every Pro and Max user on the waitlist
beta · Pro + Max · parallel threads
- 02Claude Managed Agents gets dynamic workflows, in public beta
public beta · multi-agent · API
- 03Codex now predicts your next message
beta · Pro · Tab to accept
- 04Codex on Windows gets a new sandbox built on Microsoft containers
Windows 11 · MXC sandbox
- 05Qwen-Image-2.1-Turbo makes and edits images in 8 steps
8 steps · 7B · 2K images · open weights
- 06JetBrains opens Mellum2.1, a 12B model for coding agents
12B MoE · 2.5B active · Apache 2.0
- 07Claude Dashboards and Claude Motion are in beta
beta · dashboards · animation
- 08Claude Haiku 5.5 costs ten cents per million input tokens
$0.10/M in · $0.50/M out · ~75% cheaper
- 09Liquid AI opens d1, decision models that answer in one pass
3B + 600M · one pass · open weights
- 10GPT-6 with Intelligent UI rolls out in ChatGPT
all paid plans · Free next · interactive answers
- 11SynthID Detector is open to everyone
free · image · video · audio · OpenAI + Nvidia
- 12Perplexity releases pplx-embed-v2-late retrieval models
9B + 0.6B · text + images · Hugging Face
- 13Google DeepMind ships Nano Banana 2.1 for image editing
image edit · Gemini 3.6 Flash · Gemini API
- 14EmbeddingGemma 2 brings multimodal search to phones
740M · on-device · 5 modalities · Apache 2.0
- 15Perplexity halves the price of its decision model
27B · $0.02/M in · half of v1
- 16Cursor lets you run your computer's agents from your phone
iOS · remote agents
- 17Reflection announces Beam, a 501B open-weight model
501B · 23B active · weights later in Oct
- 18Amazon Nova 2.5 Sonic is generally available
speech-to-speech · GA in Bedrock · 4 regions
- 19Reka previews Rho-1, one model for text, video and actions
19B · video + actions · preview
- 20Aleph Alpha releases Kolibri, an English-German open model
78B · 3B active · 1M context · Apache 2.0
- 21Ai2 open-sources AstaBrief for cited research reports
8B · cited reports · 51s each
- 22Whistle fits speech recognition into 16.9 MB
16.9 MB · 7 languages · runs on CPU
- 23Microsoft launches its first streaming transcription model
60 languages · streaming · voice in 23
- 24Cloudflare launches Clef decision models on Workers AI
27B + 9B · 64K context · Apache 2.0
- 25Decagon's Voice 3 runs on its own speech model, Chord
70+ languages · own speech model
- 26Tavus previews Griffin, real-time video for AI agents
0.43s latency · video-to-video · preview
This week
20 models shipped.
- 3Fri 2: AstaBrief 8B, Whistle, Ling 3.1 Flash
- 1Sat 3: Kolibri
- Sun 4: none
- 4Mon 5: Nova 2.5 Sonic, Beam, Rho-1, Hy Image 3.5
- 4Tue 6: Mistral Large 4, Nano Banana 2.1, EmbeddingGemma 2, pplx-decider 1.1
- 4Wed 7: Claude Haiku 5.5, GLM 5.3 Fast, Liquid d1, pplx-embed v2
- 3Thu 8: Mellum2.1, Grok Imagine Video 1.5 Lite, Step 5 Preview
- 1Fri 9: Qwen-Image-2.1-Turbo
- Fri 2
- Sat 3
- Sun 4
- Mon 5
- Tue 6
- Wed 7
- Thu 8
- Fri 9
Live
Straight from the labs.
All 63- We’re beginning a process of publishing more frequent reports on model behavior, beyond what appears in our system cards and regular risk reports.
- Fresh updates for dots. First up: you can now create your dot straight from your phone in the ChatGPT app on iOS and Android.
- Now in beta: composer predictions in Codex for Pro users.
- We just let in every Pro and Max user from the Claude Code Projects waitlist!
- Claude Managed Agents dynamic workflows are now available in public beta. It's a new type of multiagent orchestration, built for your most ambitious…
- An update for builders using Codex on Windows: We’ve built a new sandbox mode using @Microsoft’s Execution Containers (MXC) for faster setup,…
- 🚀 Meet Qwen-Image-2.1-Turbo — create and edit images in just 8 denoising steps! Open weights now available!
- A week of free access to Qwen3.8-Max, Qwen3.8-Flash, and Wan3.0 on @gmi_cloudis live. Explore what you can build with us now. 🙌
Spotted
What people are building with AI this week.
Models
Who is ahead, and what it costs.
Top model
Epoch Capabilities Index
Claude Opus 5.5
167.3 · Anthropic
Models tracked
274 with independent scores
337
Best open weights
#15 overall
Kimi K3
Cheapest in the top 25
GPT-6 Luna · per 1M input tokens
$0.10
“The two biggest releases this week were about cost. A trillion-parameter model that only wakes 52 billion at a time, and a Haiku at ten cents.”
Every morning by 8am Eastern.
One page, checked against the source. No account, no inbox. Follow it in any reader.