The Inference Report
The labs are converging on a single operational problem: how to make agents and long-context models run efficiently in production without collapsing under their own infrastructure costs. OpenAI is demonstrating concrete ROI through Proaction's 60% sales boost and 75-hour time savings, which signals confidence that Codex and its successors have crossed into genuine business utility. GitHub's Copilot canvas layer removes the friction between intention and interface, letting users describe workflows in plain English rather than fighting tool design. But the real pressure point is visible in AMD's three separate announcements on quantization, KV cache optimization, and weight profiling. These are not theoretical exercises. When agent prompts reach hundreds of thousands of tokens, compute stops being the bottleneck and memory becomes the wall. AMD is essentially publishing a map of where every parameter lives in a model and how to shrink it without losing reasoning accuracy. Local quantization on Strix Halo and UltraQuant's throughput gains on Instinct GPUs are not marketing moves; they are responses to a hard economic fact that nobody is saying aloud: the decode phase of agentic inference is currently expensive enough to threaten unit economics at scale. Anthropic's announcement on long-context reasoning completes the picture. The labs have collectively identified that multi-turn agents with long shared prefixes and short generations are hitting memory walls hardest, and they are racing to solve it through different angles. OpenAI is selling the outcome. GitHub is selling the interface. AMD is selling the infrastructure to make it affordable. The money is moving toward whoever can make the KV cache problem disappear first.
Sloane Duvall
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.
AllenAINone
- AllenAI: Olmo 3 32B ThinkText66K
Amazon- Amazon: Nova 2 LiteTextVisionVideoFiles1M
- Amazon: Nova Premier 1.0TextVision1M
None
Anthropic- Anthropic: Claude Fable 5TextVisionFiles1M
- Anthropic: Claude Fable 5.1TextVisionFiles1M
- Anthropic: Claude Haiku 4.5TextVisionFiles200K
- Anthropic: Claude Opus 4.5FilesVisionText200K
- Anthropic: Claude Opus 4.6TextVisionFiles1M
- Anthropic: Claude Opus 4.7TextVisionFiles1M
- Anthropic: Claude Opus 4.8TextVisionFiles1M
- Anthropic: Claude Opus 5TextVisionFiles1M
- Anthropic: Claude Opus 5.5TextVisionFiles1M
- Anthropic: Claude Sonnet 4.5TextVisionFiles1M
- Anthropic: Claude Sonnet 4.6TextVisionFiles1M
- Anthropic: Claude Sonnet 5TextVisionFiles1M
None
Arcee AINone
- Arcee AI: Trinity Large ThinkingText262K
Deep CogitoNone
- Deep Cogito: Cogito v2.1 671BText128K
Google- Google: Gemini 3 Flash PreviewTextVisionFilesAudioVideo1M
- Google: Gemini 3.1 Flash LiteTextVisionVideoFilesAudio1M
- Google: Gemini 3.1 Flash Lite PreviewTextVisionVideoFilesAudio1M
- Google: Gemini 3.1 Pro PreviewAudioFilesVisionTextVideo1M
- Google: Gemini 3.1 Pro Preview Custom ToolsTextAudioVisionVideoFiles1M
- Google: Gemini 3.5 FlashTextVisionVideoFilesAudio1M
- Google: Gemini 3.5 Flash LiteTextVisionVideoFilesAudio1M
- Google: Gemini 3.6 FlashTextVisionVideoFilesAudio1M
- Google: Gemini 3.7 FlashTextVisionVideoFilesAudio1M
- Google: Gemini 3.8 FlashTextVisionVideoFilesAudio1M
- Google: Lyria 3 Clip PreviewTextVision1M
- Google: Lyria 3 Pro PreviewTextVision1M
- 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
- Google: Gemma 4 26B A4B VisionTextVideo262K
- Google: Gemma 4 31BVisionTextVideo262K
IBMNone
- IBM: Granite 4.0 MicroText131K
- IBM: Granite 4.2 8BText131K
Inception- Inception: Mercury 2Text128K
- Inception: Mercury 2.5Text260K
None
Meta (Muse)- Meta: Muse Spark 1.1TextVisionVideoFilesAudio1M
- Meta: Muse Spark 1.2TextVisionVideoFilesAudio1M
- Meta: Muse Spark 1.2 ContributorTextVisionVideoFilesAudio1M
- Meta: Muse Spark 1.3TextVisionVideoFilesAudio1M
- Meta: Muse Spark 1.3 ContributorTextVisionVideoFilesAudio1M
- Meta: Muse Glimmer 30BTextVision131K
NVIDIANone
- NVIDIA: Nemotron 3 Nano 30B A3BText262K
- NVIDIA: Nemotron 3 SuperText262K
- NVIDIA: Nemotron 3 UltraText262K
- NVIDIA: Nemotron 3.5 LightningText1M
OpenAI- OpenAI: GPT AudioTextAudio128K
- OpenAI: GPT Audio MiniTextAudio128K
- OpenAI: GPT-5 ImageVisionTextFiles400K
- OpenAI: GPT-5 Image MiniFilesVisionText400K
- OpenAI: GPT-5 ProVisionTextFiles400K
- OpenAI: GPT-5.1VisionTextFiles400K
- OpenAI: GPT-5.1-CodexTextVision400K
- OpenAI: GPT-5.1-Codex-MaxTextVision400K
- OpenAI: GPT-5.1-Codex-MiniVisionText400K
- OpenAI: GPT-5.2FilesVisionText400K
- OpenAI: GPT-5.2 ChatFilesVisionText128K
- OpenAI: GPT-5.2 ProVisionTextFiles400K
- OpenAI: GPT-5.2-CodexTextVision400K
- OpenAI: GPT-5.3-CodexTextVisionFiles400K
- OpenAI: GPT-5.4TextVisionFiles1M
- OpenAI: GPT-5.4 Image 2VisionTextFiles272K
- OpenAI: GPT-5.4 MiniFilesVisionText400K
- OpenAI: GPT-5.4 NanoFilesVisionText400K
- OpenAI: GPT-5.4 ProTextVisionFiles1M
- OpenAI: GPT-5.5FilesVisionText1M
- OpenAI: GPT-5.5 ProFilesVisionText1M
- OpenAI: GPT-5.6 LunaFilesVisionText1M
- OpenAI: GPT-5.6 Luna ProFilesVisionText1M
- OpenAI: GPT-5.6 SolFilesVisionText1M
- OpenAI: GPT-5.6 Sol ProFilesVisionText1M
- OpenAI: GPT-5.6 TerraFilesVisionText1M
- OpenAI: GPT-5.6 Terra ProFilesVisionText1M
- OpenAI: GPT-6 AstraFilesVisionText1M
- OpenAI: GPT-6 Astra ProFilesVisionText1M
- OpenAI: GPT-6 LunaFilesVisionText1M
- OpenAI: GPT-6 Luna ProFilesVisionText1M
- OpenAI: GPT-6 SolFilesVisionText1M
- OpenAI: GPT-6 Sol ProFilesVisionText1M
None
Perceptron- Perceptron: Perceptron Mk1TextVisionVideo33K
- Perceptron: Perceptron Mk1.5TextVisionVideoAudio37K
None
Perplexity- Perplexity: Sonar Pro SearchTextVision200K
None
PoolsideNone
- Poolside: Laguna S 2.1Text1M
- Poolside: Laguna XS 2.1Text262K
Reka AINone
- Reka EdgeVisionTextVideo16K
RelaceNone
- Relace: Relace Apply 3Text256K
- Relace: Relace SearchText256K
Thinking MachinesNone
- Thinking Machines: InklingTextVisionAudio524K
- Thinking Machines: Inkling SmallTextVisionAudio524K
xAI- SpaceXAI: Grok 4.20TextVisionFiles2M
- SpaceXAI: Grok 4.20 Multi-AgentTextVisionFiles2M
- SpaceXAI: Grok 4.3TextVisionFiles1M
- SpaceXAI: Grok 4.5TextVisionFiles500K
- SpaceXAI: Grok 4.6TextVisionFiles500K
- SpaceXAI: Grok 4.7TextVisionFiles500K
- SpaceXAI: Grok Build 0.1TextVisionFiles256K
None
Cohere- Cohere: Command A+TextVision192K
None
Mistral AINone
- Mistral: Devstral 2 2512TextFiles262K
- Mistral: Ministral 3 14B 2512TextVision262K
- Mistral: Ministral 3 3B 2512TextVision131K
- Mistral: Ministral 3 8B 2512TextVision262K
- Mistral: Mistral Large 3 2512TextVisionFiles262K
- Mistral: Mistral Medium 3.5TextVisionFiles262K
- Mistral: Mistral Small 4TextVision262K
- Mistral: Voxtral Small 24B 2507TextAudioFiles33K
Alibaba (Qwen)- Qwen: Qwen3 Max ThinkingText262K
- Qwen: Qwen3.6 FlashTextVisionVideo1M
- Qwen: Qwen3.6 Max PreviewText262K
- Qwen: Qwen3.6 PlusTextVisionVideo1M
- Qwen: Qwen3.7 FlashTextVisionVideo1M
- Qwen: Qwen3.7 MaxText1M
- Qwen: Qwen3.7 PlusTextVision1M
- Qwen: Qwen3 Coder NextText262K
- Qwen: Qwen3 VL 30B A3B InstructTextVision262K
- Qwen: Qwen3 VL 30B A3B ThinkingTextVision262K
- Qwen: Qwen3 VL 32B InstructTextVision131K
- Qwen: Qwen3 VL 8B InstructVisionText262K
- Qwen: Qwen3 VL 8B ThinkingVisionText131K
- Qwen: Qwen3.5 397B A17BTextVisionVideo262K
- Qwen: Qwen3.5-122B-A10BTextVisionVideo262K
- Qwen: Qwen3.5-27BTextVisionVideo262K
- Qwen: Qwen3.5-35B-A3BTextVisionVideo262K
- Qwen: Qwen3.5-9BTextVisionVideo262K
- Qwen: Qwen3.6 27BTextVisionVideo262K
- Qwen: Qwen3.6 35B A3BTextVisionVideo262K
- Qwen: Qwen3.8 2.4T A95BText1M
- Qwen: Qwen3.8 27BTextVisionVideo1M
- Qwen: Qwen3.8 FlashTextVisionVideo1M
- Qwen: Qwen3.8 Max (0902)TextVisionVideo1M
- Qwen: Qwen3.8 Max PrimeTextVisionVideo1M
- Qwen: Qwen3.8 Omni FlashTextVisionAudioVideo1M
ByteDance- Seed: Seed 1.6VisionTextVideo262K
- Seed: Seed 1.6 FlashVisionTextVideo262K
- Seed: Seed 2.1 TurboTextVisionVideo262K
- Seed: Seed-2.0-CodeTextVisionVideo262K
- Seed: Seed-2.0-LiteTextVisionVideo262K
- Seed: Seed-2.0-MiniTextVisionVideo262K
None
DeepSeekNone
- DeepSeek: DeepSeek V3.2Text164K
- DeepSeek: DeepSeek V3.2 ExpText164K
- DeepSeek: DeepSeek V4 Flash 0423Text1M
- DeepSeek: DeepSeek V4 Flash 0731Text1M
- DeepSeek: DeepSeek V4 Flash Vision ExpTextVision1M
- DeepSeek: DeepSeek V4 Pro 0423Text1M
- DeepSeek: DeepSeek V4 Pro 0813Text1M
- DeepSeek: DeepSeek V4.1 FlashTextVision1M
MiniMax- MiniMax: MiniMax M2-herText66K
- MiniMax: MiniMax M2Text205K
- MiniMax: MiniMax M2.1Text205K
- MiniMax: MiniMax M2.5Text205K
- MiniMax: MiniMax M2.7Text205K
- MiniMax: MiniMax M3TextVisionVideo1M
Moonshot AINone
- MoonshotAI: Kimi K2 ThinkingText262K
- MoonshotAI: Kimi K2.5TextVision262K
- MoonshotAI: Kimi K2.6TextVision262K
- MoonshotAI: Kimi K2.7 CodeTextVision262K
- MoonshotAI: Kimi K3TextVisionVideo1M
XiaomiNone
- Xiaomi: MiMo-V2.5TextAudioVisionVideo1M
- Xiaomi: MiMo-V2.5-ProText1M
- Xiaomi: MiMo-V2.6-FlashTextVisionVideoAudio1M
- Xiaomi: MiMo-V2.6-ProTextVisionVideoAudio1M
- Xiaomi: MiMo-V2.6-Pro-UltraSpeedTextVisionVideoAudio1M
Z.aiNone
- Z.ai: GLM 4.6Text205K
- Z.ai: GLM 4.6VVisionTextVideo131K
- Z.ai: GLM 4.7Text205K
- Z.ai: GLM 4.7 FlashText200K
- Z.ai: GLM 5Text205K
- Z.ai: GLM 5 TurboText203K
- Z.ai: GLM 5.1Text205K
- Z.ai: GLM 5.2Text1M
- Z.ai: GLM 5.3Text1M
- Z.ai: GLM 5.3 FlashTextVisionVideo1M
- Z.ai: GLM 5.3 FlashXTextVisionVideo1M
- Z.ai: GLM 5.3 PrimeText1M
- Z.ai: GLM 5V TurboVisionTextVideo203K
Upstage- Upstage: Solar Mini 4Text524K
- Upstage: Solar Pro 3Text131K
- Upstage: Solar Pro 4Text524K
None