Reference Matrix · AI API Limits & Performance Matrix
AI API Limits & Performance34 model matrices, side by side
Rate limits, context windows, pricing and tier requirements for every major model — verified and kept current so you can size an API build before you write a line of code.
Building on an LLM API? Talk to us34
Models
13
Providers
4.9Clutch rating ⭐
600+Happy partners
17+Countries served
190+Team members
AI-Powered Search
Describe your problem, we'll find the right matrix
Type in plain English — no need to know the exact matrix name
Showing 34 models
| Model | Provider | Context | Status | Limits & pricing |
|---|---|---|---|---|
Yi-Largeaccounts/fireworks/models/yi-large | 01.AI | 32K–33K tokens | GA | View matrix |
Qwen2not published | Alibaba | 128K tokens (Qwen2-72B) | GA | View matrix |
Qwen3-Maxqwen3-max / qwen3-max-2026-01-23 | Alibaba | 262K tokens | GA | View matrix |
Claude 3 Haikuclaude-haiku-4-5 | Anthropic | 200K tokens | GA | View matrix |
Claude 3.5 Sonnetclaude-sonnet-5 | Anthropic | 1M tokens | GA | View matrix |
Claude 4 Opusclaude-opus-4-8 | Anthropic | 1M tokens (no surcharge) | GA | View matrix |
Claude 4 Sonnetclaude-sonnet-5 | Anthropic | 1M tokens | GA | View matrix |
Cohere Command Rcommand-r-08-2024 | Cohere | 128K tokens | GA | View matrix |
Gemini 1.5 Flashgemini-1.5-flash | 1M tokens (retired) | Deprecated | View matrix | |
Gemini 1.5 Progemini-1.5-pro | 1M+ tokens (was the headline feature at launch) | Deprecated | View matrix | |
Gemini 2.5 Flashgemini-2.5-flash | 1M tokens | Live | View matrix | |
Gemini 2.5 Progemini-2.5-pro | 1M+ tokens | Live | View matrix | |
Gemma 2gemma-2 | 8K tokens (2B, 9B, 27B variants) | Deprecated | View matrix | |
Gemma 3gemma-3-4b-it | 131K tokens | Live | View matrix | |
Imagen 4imagen-4.0-generate-001 | n/a (image prompt) | Deprecated | View matrix | |
Llama 3meta-llama/Llama-3-8b-instruct | Meta | 8K–128K tokens (host-dependent) | GA | View matrix |
Llama 3.1meta-llama/Llama-3.1-70B-Instruct | Meta | 128K tokens | GA | View matrix |
Llama 4 Maverickmeta-llama/Llama-4-Maverick-17B-128E-Instruct | Meta | 1M tokens (native); host-capped (Groq narrower; Fireworks ~1M) | GA | View matrix |
Llama 4 Scoutmeta-llama/Llama-4-Scout-17B-16E-Instruct | Meta | 10M tokens (native); 128K (Groq), 320K (DeepInfra), ~1M (Fireworks); full 10M only via self-hosting | GA | View matrix |
Phi-3Phi-3-mini-4k-instruct / Phi-3-medium-128k-instruct | Microsoft | 4K–128K tokens (depending on Phi-3 variant) | Deprecated | View matrix |
Devstraldevstral | Mistral AI | 256K tokens | GA | View matrix |
Mistral Largemistral-large-3 | Mistral AI | 128K tokens | GA | View matrix |
Mistral Mediummistral-medium-3.5 | Mistral AI | 128K tokens | GA | View matrix |
Mixtral 8x22Bopen-mixtral-8x22b | Mistral AI | 65K tokens | GA | View matrix |
Mixtral 8x7Bopen-mixtral-8x7b | Mistral AI | 32K tokens | GA | View matrix |
Kimi K2kimi-k2.6 | Moonshot AI | 262K tokens (K2.6); 256K tokens (K2.7 Code) | GA | View matrix |
GPT-3.5 Turbogpt-3.5-turbo | OpenAI | 16K tokens (gpt-3.5-turbo-16k) / 4K tokens (base) | GA | View matrix |
GPT-4 Turbogpt-4-turbo | OpenAI | 128K tokens | GA | View matrix |
GPT-4ogpt-4o | OpenAI | 128K tokens | GA | View matrix |
GPT-5gpt-5.5 | OpenAI | ~1M tokens (922K input / 128K output) | GA | View matrix |
Stable Virtual Cameranot applicable | Stability AI | n/a | Preview | View matrix |
Falcon 180Bnot applicable | TII | not published (2K-4K trained; some sources cite 8K — verify with TII before relying on this figure) | GA | View matrix |
Grok 2grok-2 / grok-2-vision | xAI | not published for Grok 2 (retired); Grok 4.5 successor: 500K tokens | Deprecated | View matrix |
Grok 4grok-4.5 | xAI | 500K tokens (requests >200K may bill at a higher unpublished rate) | Deprecated | View matrix |
Yi-Large
GA01.AI·32K–33K tokens context
Qwen2
GAAlibaba·128K tokens (Qwen2-72B) context
Qwen3-Max
GAAlibaba·262K tokens context
Claude 3 Haiku
GAAnthropic·200K tokens context
Claude 3.5 Sonnet
GAAnthropic·1M tokens context
Claude 4 Opus
GAAnthropic·1M tokens (no surcharge) context
Claude 4 Sonnet
GAAnthropic·1M tokens context
Cohere Command R
GACohere·128K tokens context
Gemini 1.5 Flash
DeprecatedGoogle·1M tokens (retired) context
Gemini 1.5 Pro
DeprecatedGoogle·1M+ tokens (was the headline feature at launch) context
Gemini 2.5 Flash
LiveGoogle·1M tokens context
Gemini 2.5 Pro
LiveGoogle·1M+ tokens context
Gemma 2
DeprecatedGoogle·8K tokens (2B, 9B, 27B variants) context
Gemma 3
LiveGoogle·131K tokens context
Imagen 4
DeprecatedGoogle·n/a (image prompt) context
Llama 3
GAMeta·8K–128K tokens (host-dependent) context
Llama 3.1
GAMeta·128K tokens context
Llama 4 Maverick
GAMeta·1M tokens (native); host-capped (Groq narrower; Fireworks ~1M) context
Llama 4 Scout
GAMeta·10M tokens (native); 128K (Groq), 320K (DeepInfra), ~1M (Fireworks); full 10M only via self-hosting context
Phi-3
DeprecatedMicrosoft·4K–128K tokens (depending on Phi-3 variant) context
Devstral
GAMistral AI·256K tokens context
Mistral Large
GAMistral AI·128K tokens context
Mistral Medium
GAMistral AI·128K tokens context
Mixtral 8x22B
GAMistral AI·65K tokens context
Mixtral 8x7B
GAMistral AI·32K tokens context
Kimi K2
GAMoonshot AI·262K tokens (K2.6); 256K tokens (K2.7 Code) context
GPT-3.5 Turbo
GAOpenAI·16K tokens (gpt-3.5-turbo-16k) / 4K tokens (base) context
GPT-4 Turbo
GAOpenAI·128K tokens context
GPT-4o
GAOpenAI·128K tokens context
GPT-5
GAOpenAI·~1M tokens (922K input / 128K output) context
Stable Virtual Camera
PreviewStability AI·n/a context
Falcon 180B
GATII·not published (2K-4K trained; some sources cite 8K — verify with TII before relying on this figure) context
Grok 2
DeprecatedxAI·not published for Grok 2 (retired); Grok 4.5 successor: 500K tokens context
Grok 4
DeprecatedxAI·500K tokens (requests >200K may bill at a higher unpublished rate) context