Discover language, multimodal, image, video, audio and embedding models. Compare capabilities, context limits, platform pricing and original sources.
meta-llama/llama-4-scout-17b-16e-instruct
Llama 4 Scout 17B Instruct (16E) is a mixture-of-experts (MoE) language model developed by Meta, activating 17 billion parameters out of a total of 109B. It supports native multimodal input (text and image) and multilingual output (text and code) across 12 supported languages. De
- Context window
- 1,310,720
- Input modalities
- text, image
- Output modalities
- text
- Capabilities
- multimodal, structuredOutputs, toolCalling, visionInput
Benchmarks
- Artificial Analysis: 6.5
- BFCL: 28.13%
- LLM Stats · GitHub archive: 0.678
- LMArena: 1,279.207
Inference prices
- Published price : $0.10 / input_tokens_per_million · deepInfra ↗
- Published price : $0.30 / output_tokens_per_million · deepInfra ↗
google/gemma-3-12b-it
Gemma 3 introduces multimodality, supporting vision-language input and text outputs. It handles context windows up to 128k tokens, understands over 140 languages, and offers improved math, reasoning, and chat capabilities, including structured outputs and function calling. Gemma
- Context window
- 131,072
- Input modalities
- text, image
- Output modalities
- text
- Capabilities
- multimodal, structuredOutputs, toolCalling, visionInput
Benchmarks
- Artificial Analysis: 3.8
- LMArena: 1,334.209
- MMBench: 97.6%
- Open Persian LLM Leaderboard: 93.39%
- OpenRouter Benchmarks: 38.047%
Inference prices
mistralai/ministral-3b-2512
The smallest model in the Ministral 3 family, Ministral 3 3B is a powerful, efficient tiny language model with vision capabilities.
- Context window
- 131,072
- Input modalities
- text, image
- Output modalities
- text
- Capabilities
- batch, multimodal, promptCaching, streaming, structuredOutputs, toolCalling, visionInput
Benchmarks
Inference prices
- input / 1M tokens : $0.10 / input_tokens_per_million · mistral ↗
- output / 1M tokens : $0.10 / output_tokens_per_million · mistral ↗
google/gemma-3-4b-it
Gemma 3 introduces multimodality, supporting vision-language input and text outputs. It handles context windows up to 128k tokens, understands over 140 languages, and offers improved math, reasoning, and chat capabilities, including structured outputs and function calling.
- Context window
- 131,072
- Input modalities
- text, image
- Output modalities
- text
- Capabilities
- multimodal, structuredOutputs, toolCalling, visionInput
Benchmarks
- Artificial Analysis: 4.8
- LMArena: 1,290.792
- MMBench: 98.9%
- Open Persian LLM Leaderboard: 63.76%
- OpenRouter Benchmarks: 27.104%
Inference prices
cloudflareWorkersAI/google/gemini-3-flash
Gemini 3 Flash is Google's fast multimodal model with frontier intelligence, superior search, and grounding capabilities.
- Context window
- 1,000,000
- Maximum output
- 8,192
- Input modalities
- text
- Output modalities
- text
- Capabilities
- multimodal, promptCaching, reasoning, structuredOutputs, toolCalling, visionInput, webSearch
Benchmarks
Inference prices
cheaperInference/kimi-k2.6
- Context window
- 256,000
- Maximum output
- 262,144
- Input modalities
- text, image
- Output modalities
- text
- Capabilities
- multimodal, reasoning, structuredOutputs, toolCalling, visionInput
Benchmarks
Inference prices
vertexAI/publishers/anthropic/models/claude-opus-4-5@20251101
- Capabilities
- multimodal, reasoning, toolCalling, visionInput
Benchmarks
- BFCL: 77.47%
- LMArena: 1,451.003
- SWE-bench: 74.4%
x-ai/grok-4.20-multi-agent-20260309
Grok 4.20 Multi-Agent is a variant of SpaceXAI’s Grok 4.20 designed for collaborative, agent-based workflows. Multiple agents operate in parallel to conduct deep research, coordinate tool use, and synthesize information across complex tasks. Reasoning effort behavior: - low / me
- Context window
- 2,000,000
- Input modalities
- text, image, file
- Output modalities
- text
- Capabilities
- fileInput, multimodal, promptCaching, reasoning, structuredOutputs, toolCalling, visionInput, webSearch
Benchmarks
Inference prices
- Cached input / 1M tokens : $0.20 / cache_read_tokens_per_million · xAI ↗
- Input / 1M tokens : $1.25 / input_tokens_per_million · xAI ↗
- Output / 1M tokens : $2.50 / output_tokens_per_million · xAI ↗
vertexAI/publishers/hf-nvidia/models/nvidia-nemotron-3-ultra-550b-a55b-nvfp4@001
- Input modalities
- text
- Output modalities
- text
Benchmarks
vercelAIGatewayCatalog/alibaba/qwen3-max-preview
Qwen3-Max-Preview shows substantial gains over the 2.5 series in overall capability, with significant enhancements in Chinese-English text understanding, complex instruction following, handling of subjective open-ended tasks, multilingual ability, and tool invocation; model knowl
- Context window
- 262,144
- Maximum output
- 32,768
- Input modalities
- text
- Output modalities
- text
Benchmarks
deepseek/deepseek-v3.2-exp
DeepSeek-V3.2-Exp is an experimental large language model released by DeepSeek as an intermediate step between V3.1 and future architectures. It introduces DeepSeek Sparse Attention (DSA), a fine-grained sparse attention mechanism designed to improve training and inference effici
- Context window
- 163,840
- Input modalities
- text
- Output modalities
- text
- Capabilities
- reasoning, structuredOutputs, toolCalling
Benchmarks
- Artificial Analysis: 16.6
- BFCL: 56.73%
- LMArena: 1,422.402
Inference prices
- Published price : $0.21 / input_tokens_per_million · deepInfra ↗
- Published price : $0.32 / output_tokens_per_million · deepInfra ↗
qwen/qwen3-235b-a22b-07-25
Qwen3-235B-A22B-Instruct-2507 is a multilingual, instruction-tuned mixture-of-experts language model based on the Qwen3-235B architecture, with 22B active parameters per forward pass. It is optimized for general-purpose text generation, including instruction following, logical re
- Context window
- 262,144
- Input modalities
- text
- Output modalities
- text
- Capabilities
- promptCaching, structuredOutputs, toolCalling
Benchmarks
- Artificial Analysis: 12
- LLM Stats · GitHub archive: 0.95
- LMArena: 1,419.17
- OpenCompass LLM: 12.3%
- OpenRouter Benchmarks: 40.667%
Inference prices
- Published price : $0.05 / cache_read_tokens_per_million · openRouter ↗
- Published price : $0.14 / input_tokens_per_million · openRouter ↗
- Published price : $0.80 / output_tokens_per_million · openRouter ↗
anthropic/claude-4.1-opus-20250805
Claude Opus 4.1 is an updated version of Anthropic’s flagship model, offering improved performance in coding, reasoning, and agentic tasks. It achieves 74.5% on SWE-bench Verified and shows notable gains in multi-file code refactoring, debugging precision, and detail-oriented rea
- Context window
- 200,000
- Input modalities
- image, text, file
- Output modalities
- text
- Capabilities
- fileInput, multimodal, promptCaching, reasoning, toolCalling, visionInput
Benchmarks
- Artificial Analysis: 22.8
- LMArena: 1,418.633
Inference prices
- Published price : $1.50 / cache_read_tokens_per_million · openRouter ↗
- Published price : $30.00 / cache_write_1h_tokens_per_million · openRouter ↗
- Published price : $18.75 / cache_write_tokens_per_million · openRouter ↗
- Published price : $15.00 / input_tokens_per_million · openRouter ↗
- Published price : $75.00 / output_tokens_per_million · openRouter ↗
google/gemini-2.5-flash
Gemini 2.5 Flash is Google's state-of-the-art workhorse model, specifically designed for advanced reasoning, coding, mathematics, and scientific tasks. It includes built-in "thinking" capabilities, enabling it to provide responses with greater accuracy and nuanced context handlin
- Context window
- 1,048,576
- Input modalities
- file, image, text, audio, video
- Output modalities
- text
- Capabilities
- audioInput, batch, fileInput, multimodal, promptCaching, reasoning, structuredOutputs, toolCalling, visionInput, webSearch
Benchmarks
- Artificial Analysis: 15.5
- BFCL: 56.24%
- Kaggle Benchmarks: 50%
- LLM Stats · GitHub archive: 0.797
- LMArena: 1,416.988
- OpenRouter Benchmarks: 54.667%
- SWE-bench: 28.73%
Inference prices
openai/gpt-5.2-20251211
GPT-5.2 is the latest frontier-grade model in the GPT-5 series, offering stronger agentic and long context perfomance compared to GPT-5.1. It uses adaptive reasoning to allocate computation dynamically, responding quickly to simple queries while spending more depth on complex tas
- Context window
- 400,000
- Input modalities
- file, image, text
- Output modalities
- text
- Capabilities
- fileInput, multimodal, promptCaching, reasoning, structuredOutputs, toolCalling, visionInput, webSearch
Benchmarks
- Artificial Analysis: 30.4
- LMArena: 1,416.33
- SWE-bench: 69%
Inference prices
openai/o3-2025-04-16
o3 is a well-rounded and powerful model across domains. It sets a new standard for math, science, coding, and visual reasoning tasks. It also excels at technical writing and instruction-following. Use it to think through multi-step problems that involve analysis across text, code
- Context window
- 200,000
- Input modalities
- image, text, file
- Output modalities
- text
- Capabilities
- fileInput, multimodal, promptCaching, reasoning, structuredOutputs, toolCalling, visionInput
Benchmarks
- Artificial Analysis: 20.2
- BFCL: 63.05%
- Kaggle Benchmarks: 50%
- LLM Stats · GitHub archive: 0.868
- LMArena: 1,409.751
- SWE-bench: 58.4%
Inference prices
- Published price : $0.50 / cache_read_tokens_per_million · openRouter ↗
- Published price : $2.00 / input_tokens_per_million · openRouter ↗
- Published price : $8.00 / output_tokens_per_million · openRouter ↗
cheaperInference/grok-4.1-fast-reasoning
- Context window
- 1,000,000
- Maximum output
- 1,000,000
- Input modalities
- text, image
- Output modalities
- text
- Capabilities
- multimodal, reasoning, structuredOutputs, toolCalling, visionInput
Benchmarks
Inference prices
azureFoundry/azure-huggingface/nvidia--nvidia-nemotron-3.5-super-vl-120b-a12b-nvfp4@1
NVFP4 vision-language reasoning model for document understanding, coding, and tool use.
- Context window
- 1,048,576
- Input modalities
- text, image, video
- Output modalities
- text
- Capabilities
- multimodal, visionInput
Benchmarks