sonar-reasoning LMArena Search Arena 24K
list all of the available ai models from gork, anthropic, open ai, and deepseek
- User
list all of the available ai models from gork, anthropic, open ai, and deepseek
- Assistant
Anthropic Models
Claude 3.7 Sonnet
- Max tokens: 200K
- Languages: English, French, Arabic, Mandarin, Hindi, Spanish, Portuguese, Korean, Japanese, German, Russian, others[1].
- Capabilities: Extended thinking for complex problem-solving, agentic coding, context understanding, and creative tasks[1].
Claude 3.5 Sonnet
- Max tokens: 200K
- Languages: English, Spanish, Japanese, others[1][4].
- Capabilities: Advanced coding, multi-step workflows, visual processing, and natural writing[1][4].
Claude 3.5 Haiku
- Max tokens: 200K
- Languages: English, Spanish, Japanese, others[1][4].
- Capabilities: Fast code completions, chatbots, data extraction, and real-time content moderation[1][4].
Claude 3 Opus
- Max tokens: 200K
- Languages: English, Spanish, Japanese, others[1][4].
- Capabilities: Task automation, research analysis, financial forecasting, and vision tasks[1][4].
Claude 3 Haiku
- Max tokens: 200K
- Languages: English, Spanish, Japanese, others[1][4].
- Capabilities: Live interactions, translations, content moderation, and inventory management[1][4].
Claude 3 Sonnet
- Max tokens: 200K
- Languages: English, Spanish, Japanese, others[1].
- Capabilities: RAG, product recommendations, code generation, and image parsing[1].
OpenAI Models
GPT-4 Omni
- Capabilities: Human-like reasoning, versatility across tasks, and fast performance[3].
DeepSeek Models
DeepSeek-R1
- Focus: Advanced reasoning, matching OpenAI’s o1 in benchmarks[2][5][8].
- Variants: R1-Zero (trained via reinforcement learning)[2].
DeepSeekMoE
- Versions: V2, V2.5, V3 (671B parameters, 128K context)[2][5].
- Architecture: Mixture-of-experts for specialized tasks[2].
DeepSeek LLM
- Versions: 7B, 67B (trained on 2T tokens in English/Chinese)[2][5].
DeepSeek-Coder
- Versions: V1 (80+ languages), V2 (236B parameters, 6T tokens)[2][5].
DeepSeek-VL
- Versions: VL (500B text + 400B vision tokens), VL2 (Tiny/Small/Base)[2][5].
DeepSeek-Math
- Focus: Mathematical reasoning (120B math tokens)[2][5].
DeepSeek-Prover
- Versions: V1, V1.5 (automated theorem proving in Lean 4)[2][5].
Janus
- Variants: JanusFlow, Janus-Pro (multimodal understanding/generation)[2][5].
Grok (xAI) Models
Grok-2
- Capabilities: General-purpose tasks (excludes image generation)[3][6].
- Variant: Grok-2-1212 (specific implementation)[6].
Key Comparisons
Model Max Tokens Specialization Sources Claude 3.7 Sonnet 200K Complex reasoning, coding [1][4] DeepSeek-R1 128K Advanced reasoning, math [2][5][8] GPT-4 Omni ~32K General-purpose tasks [3] Grok-2 N/A Versatile, non-visual tasks [3][6]
Other models, same or similar prompt
Responses were produced in different, uncontrolled contexts (system prompts, settings, prior turns). This is not a benchmark.