The Inference Report

September 4, 2026

AI Labs — News

OpenAI is doubling down on model capability as the primary distribution channel, using GPT-6 Astra's performance gains in document review and game prototyping to justify enterprise adoption while simultaneously positioning itself as infrastructure for national security through the Daybreak program. The move signals confidence that raw capability, not regulatory approval or safety certifications, drives adoption; the Safety overview mentions Critical-level cybersecurity capability almost as an afterthought to the business cases. Meanwhile, NVIDIA's acquisition of Hugging Face for $12.93 billion consolidates control over the open-source model distribution layer just as that layer matters most. NVIDIA gains direct ownership of the platform where developers discover, fine-tune, and deploy models, turning what was a neutral commons into a vertically integrated advantage. AMD's parallel announcement that DeepSeek-V4-Flash runs on MI355X GPUs suggests the real competition isn't between labs but between hardware vendors fighting for training and inference workloads; the open-weight model becomes the battleground, not the prize. Google and DeepMind's announcements, genomic prediction, fruit fly connectomics, WeatherNext 3, occupy a different lane entirely, emphasizing scientific application and specialized capability rather than general-purpose deployment. GitHub's parallel agents feature and Hugging Face's work on memory-owning agents and efficient structured outputs all point toward the same conclusion: the industry is betting that value accrues to whoever controls the tooling layer where models get deployed, not to whoever trains the largest weights. The real story isn't which lab released what model; it's that infrastructure and distribution are consolidating while capability becomes commoditized across an expanding range of specialized tasks.

Sloane Duvall

AMD

Enabling DeepSeek-V4-Flash Training on AMD Instinct MI355X GPUs with Primus

GitHub Blog

GitHub Copilot app for Beginners: Run several agents at once

Google

Google DeepMind

Introducing WeatherNext 3, our most advanced and accurate global weather AI model

Hugging Face

NVIDIA

OpenAI

AI Labs — Models

A curated reference of models from major AI labs, with open/closed weight status, input modalities, and context window size. American labs tend towards closed weights models and Chinese labs tend toward open weights models.

AllenAI

Closed Weights

None

Open Weights

AllenAI: Olmo 3 32B Think
Text66K

Amazon

Closed Weights

Amazon: Nova 2 Lite
TextVisionVideoFiles1M
Amazon: Nova Premier 1.0
TextVision1M

Open Weights

None

Anthropic

Closed Weights

Anthropic: Claude Fable 5
TextVisionFiles1M
Anthropic: Claude Fable 5.1
TextVisionFiles1M
Anthropic: Claude Haiku 4.5
TextVisionFiles200K
Anthropic: Claude Opus 4.5
FilesVisionText200K
Anthropic: Claude Opus 4.6
TextVisionFiles1M
Anthropic: Claude Opus 4.7
TextVisionFiles1M
Anthropic: Claude Opus 4.8
TextVisionFiles1M
Anthropic: Claude Sonnet 4.5
TextVisionFiles1M
Anthropic: Claude Sonnet 4.6
TextVisionFiles1M
Anthropic: Claude Sonnet 5
TextVisionFiles1M
Claude Opus 5
TextVisionFiles1M

Open Weights

None

Arcee AI

Closed Weights

None

Open Weights

Arcee AI: Trinity Large Thinking
Text262K

Deep Cogito

Closed Weights

None

Open Weights

Deep Cogito: Cogito v2.1 671B
Text128K

Google

Closed Weights

Google: Gemini 3 Flash Preview
TextVisionFilesAudioVideo1M
Google: Gemini 3.1 Flash Lite
TextVisionVideoFilesAudio1M
Google: Gemini 3.1 Flash Lite Preview
TextVisionVideoFilesAudio1M
Google: Gemini 3.1 Pro Preview
AudioFilesVisionTextVideo1M
Google: Gemini 3.1 Pro Preview Custom Tools
TextAudioVisionVideoFiles1M
Google: Gemini 3.5 Flash
TextVisionVideoFilesAudio1M
Google: Gemini 3.5 Flash Lite
TextVisionVideoFilesAudio1M
Google: Gemini 3.6 Flash
TextVisionVideoFilesAudio1M
Google: Gemini 3.7 Flash
TextVisionVideoFilesAudio1M
Google: Gemini 3.8 Flash
TextVisionVideoFilesAudio1M
Google: Lyria 3 Clip Preview
TextVision1M
Google: Lyria 3 Pro Preview
TextVision1M
Google: Nano Banana (Gemini 2.5 Flash Image)
VisionText33K
Google: Nano Banana 2 (Gemini 3.1 Flash Image Preview)
VisionText66K
Google: Nano Banana 2 (Gemini 3.1 Flash Image)
VisionText131K
Google: Nano Banana 2 Lite (Gemini 3.1 Flash Lite Image)
VisionText66K
Google: Nano Banana Pro (Gemini 3 Pro Image Preview)
VisionText66K
Google: Nano Banana Pro (Gemini 3 Pro Image)
VisionText131K

Open Weights

Google: Gemma 4 26B A4B
VisionTextVideo262K
Google: Gemma 4 31B
VisionTextVideo262K

IBM

Closed Weights

None

Open Weights

IBM: Granite 4.0 Micro
Text131K
IBM: Granite 4.2 8B
Text131K

Inception

Closed Weights

Inception: Mercury 2
Text128K
Inception: Mercury 2.5 Preview
Text260K

Open Weights

None

Meta (Muse)

Closed Weights

Meta: Muse Spark 1.1
TextVisionVideoFilesAudio1M
Meta: Muse Spark 1.2
TextVisionVideoFilesAudio1M
Meta: Muse Spark 1.2 Contributor
TextVisionVideoFilesAudio1M
Meta: Muse Spark 1.3
TextVisionVideoFilesAudio1M
Meta: Muse Spark 1.3 Contributor
TextVisionVideoFilesAudio1M

Open Weights

Meta: Muse Glimmer 30B
TextVision131K

NVIDIA

Closed Weights

None

Open Weights

NVIDIA: Nemotron 3 Nano 30B A3B
Text262K
NVIDIA: Nemotron 3 Super
Text1M
NVIDIA: Nemotron 3 Ultra
Text262K
NVIDIA: Nemotron 3.5 Lightning
Text262K

OpenAI

Closed Weights

OpenAI: GPT Audio
TextAudio128K
OpenAI: GPT Audio Mini
TextAudio128K
OpenAI: GPT-5 Image
VisionTextFiles400K
OpenAI: GPT-5 Image Mini
FilesVisionText400K
OpenAI: GPT-5 Pro
VisionTextFiles400K
OpenAI: GPT-5.1
VisionTextFiles400K
OpenAI: GPT-5.1-Codex
TextVision400K
OpenAI: GPT-5.1-Codex-Max
TextVision400K
OpenAI: GPT-5.1-Codex-Mini
VisionText400K
OpenAI: GPT-5.2
FilesVisionText400K
OpenAI: GPT-5.2 Chat
FilesVisionText128K
OpenAI: GPT-5.2 Pro
VisionTextFiles400K
OpenAI: GPT-5.2-Codex
TextVision400K
OpenAI: GPT-5.3-Codex
TextVisionFiles400K
OpenAI: GPT-5.4
TextVisionFiles1M
OpenAI: GPT-5.4 Image 2
VisionTextFiles272K
OpenAI: GPT-5.4 Mini
FilesVisionText400K
OpenAI: GPT-5.4 Nano
FilesVisionText400K
OpenAI: GPT-5.4 Pro
TextVisionFiles1M
OpenAI: GPT-5.5
FilesVisionText1M
OpenAI: GPT-5.5 Pro
FilesVisionText1M
OpenAI: GPT-5.6 Luna
FilesVisionText1M
OpenAI: GPT-5.6 Luna Pro
FilesVisionText1M
OpenAI: GPT-5.6 Sol
FilesVisionText1M
OpenAI: GPT-5.6 Sol Pro
FilesVisionText1M
OpenAI: GPT-5.6 Terra
FilesVisionText1M
OpenAI: GPT-5.6 Terra Pro
FilesVisionText1M
OpenAI: GPT-6 Astra
FilesVisionText1M
OpenAI: GPT-6 Astra Pro
FilesVisionText1M

Open Weights

None

Perceptron

Closed Weights

Perceptron: Perceptron Mk1
TextVisionVideo33K

Open Weights

None

Perplexity

Closed Weights

Perplexity: Sonar Pro Search
TextVision200K

Open Weights

None

Poolside

Closed Weights

None

Open Weights

Poolside: Laguna S 2.1
Text1M
Poolside: Laguna XS 2.1
Text262K

Reka AI

Closed Weights

None

Open Weights

Reka Edge
VisionTextVideo16K

Relace

Closed Weights

None

Open Weights

Relace: Relace Apply 3
Text256K
Relace: Relace Search
Text256K

Thinking Machines

Closed Weights

None

Open Weights

Thinking Machines: Inkling
TextVisionAudio1M
Thinking Machines: Inkling Small
TextVisionAudio1M

xAI

Closed Weights

SpaceXAI: Grok 4.20
TextVisionFiles2M
SpaceXAI: Grok 4.20 Multi-Agent
TextVisionFiles2M
SpaceXAI: Grok 4.3
TextVisionFiles1M
SpaceXAI: Grok 4.5
TextVisionFiles500K
SpaceXAI: Grok 4.6
TextVisionFiles500K
SpaceXAI: Grok Build 0.1
TextVisionFiles256K

Open Weights

None

Mistral AI

Closed Weights

None

Open Weights

Mistral: Devstral 2 2512
TextFiles262K
Mistral: Ministral 3 14B 2512
TextVision262K
Mistral: Ministral 3 3B 2512
TextVision131K
Mistral: Ministral 3 8B 2512
TextVision262K
Mistral: Mistral Large 3 2512
TextVisionFiles262K
Mistral: Mistral Medium 3.5
TextVisionFiles262K
Mistral: Mistral Small 4
TextVision262K
Mistral: Voxtral Small 24B 2507
TextAudioFiles33K

Alibaba (Qwen)

Closed Weights

Qwen: Qwen3 Coder Flash
Text1M
Qwen: Qwen3 Coder Plus
Text1M
Qwen: Qwen3 Max
Text262K
Qwen: Qwen3 Max Thinking
Text262K
Qwen: Qwen3.6 Flash
TextVisionVideo1M
Qwen: Qwen3.6 Max Preview
Text262K
Qwen: Qwen3.6 Plus
TextVisionVideo1M
Qwen: Qwen3.7 Flash
TextVisionVideo1M
Qwen: Qwen3.7 Max
Text1M
Qwen: Qwen3.7 Plus
TextVision1M

Open Weights

Qwen: Qwen3 Coder Next
Text262K
Qwen: Qwen3 Next 80B A3B Instruct
Text262K
Qwen: Qwen3 Next 80B A3B Thinking
Text262K
Qwen: Qwen3 VL 235B A22B Instruct
TextVision262K
Qwen: Qwen3 VL 235B A22B Thinking
TextVision131K
Qwen: Qwen3 VL 30B A3B Instruct
TextVision262K
Qwen: Qwen3 VL 30B A3B Thinking
TextVision262K
Qwen: Qwen3 VL 32B Instruct
TextVision131K
Qwen: Qwen3 VL 8B Instruct
VisionText262K
Qwen: Qwen3 VL 8B Thinking
VisionText131K
Qwen: Qwen3.5 397B A17B
TextVisionVideo262K
Qwen: Qwen3.5-122B-A10B
TextVisionVideo262K
Qwen: Qwen3.5-27B
TextVisionVideo262K
Qwen: Qwen3.5-35B-A3B
TextVisionVideo262K
Qwen: Qwen3.5-9B
TextVisionVideo262K
Qwen: Qwen3.6 27B
TextVisionVideo262K
Qwen: Qwen3.6 35B A3B
TextVisionVideo262K
Qwen: Qwen3.8 2.4T A95B
Text1M
Qwen: Qwen3.8 27B
TextVisionVideo1M
Qwen: Qwen3.8 Flash
TextVisionVideo1M
Qwen: Qwen3.8 Max (0902)
TextVisionVideo1M

ByteDance

Closed Weights

Seed: Seed 1.6
VisionTextVideo262K
Seed: Seed 1.6 Flash
VisionTextVideo262K
Seed: Seed 2.1 Turbo
TextVisionVideo262K
Seed: Seed-2.0-Code
TextVisionVideo262K
Seed: Seed-2.0-Lite
TextVisionVideo262K
Seed: Seed-2.0-Mini
TextVisionVideo262K

Open Weights

None

DeepSeek

Closed Weights

None

Open Weights

DeepSeek: DeepSeek V3.1 Terminus
Text164K
DeepSeek: DeepSeek V3.2
Text164K
DeepSeek: DeepSeek V3.2 Exp
Text164K
DeepSeek: DeepSeek V4 Flash 0423
Text1M
DeepSeek: DeepSeek V4 Flash 0731
Text1M
DeepSeek: DeepSeek V4 Flash Vision Exp
TextVision1M
DeepSeek: DeepSeek V4 Pro 0423
Text1M
DeepSeek: DeepSeek V4 Pro 0813
Text1M

MiniMax

Closed Weights

MiniMax: MiniMax M2-her
Text66K

Open Weights

MiniMax: MiniMax M2
Text205K
MiniMax: MiniMax M2.1
Text205K
MiniMax: MiniMax M2.5
Text205K
MiniMax: MiniMax M2.7
Text205K
MiniMax: MiniMax M3
TextVisionVideo1M

Moonshot AI

Closed Weights

None

Open Weights

MoonshotAI: Kimi K2 Thinking
Text262K
MoonshotAI: Kimi K2.5
TextVision262K
MoonshotAI: Kimi K2.6
TextVision262K
MoonshotAI: Kimi K2.7 Code
TextVision262K
MoonshotAI: Kimi K3
TextVisionVideo1M

Xiaomi

Closed Weights

None

Open Weights

Xiaomi: MiMo-V2.5
TextAudioVisionVideo1M
Xiaomi: MiMo-V2.5-Pro
Text1M

Z.ai

Closed Weights

None

Open Weights

Z.ai: GLM 4.6
Text205K
Z.ai: GLM 4.6V
VisionTextVideo131K
Z.ai: GLM 4.7
Text205K
Z.ai: GLM 4.7 Flash
Text203K
Z.ai: GLM 5
Text205K
Z.ai: GLM 5 Turbo
Text203K
Z.ai: GLM 5.1
Text205K
Z.ai: GLM 5.2
Text1M
Z.ai: GLM 5.3
Text1M
Z.ai: GLM 5.3 Flash
TextVisionVideo1M
Z.ai: GLM 5V Turbo
VisionTextVideo203K

Upstage

Closed Weights

Upstage: Solar Pro 3
Text131K
Upstage: Solar Pro 4
Text524K

Open Weights

None