196 results
Foundation models you can call or deploy
by Mistral AI·Sep 2023
Mistral's 7.3B dense open-weight LLM under Apache 2.0; uses grouped-query and sliding-window attention, a widely fine-tuned baseline for local inference.
by Anthropic·Feb 2025
Anthropic's hybrid-reasoning Claude model with toggleable extended thinking; reached state-of-the-art scores on SWE-bench Verified at launch.
by OpenAI·Dec 2024
Higher-compute variant of OpenAI's o1 reasoning model, exclusive to ChatGPT Pro subscribers; uses more inference-time compute for harder math and coding problems.
by OpenAI·Apr 2025
OpenAI's API-only GPT-4 update with improved instruction following, coding gains, and a 1M-token context window; positioned as a cost-efficient GPT-4o alternative.
by OpenAI·Mar 2023
OpenAI's earlier flagship multimodal (text+image) LLM, predecessor to GPT-4o, with 8K/32K context windows and strong performance on professional and academic benchmarks.
by OpenAI·Jan 2025
Smaller, faster, cheaper reasoning variant of o3 with three selectable effort levels (low/medium/high); released as a successor to o1-mini for cost-sensitive workloads.
by DeepSeek·Dec 2024
671B/37B-active MoE with MLA and MTP, 128K context, MIT license; trained on 14.8T tokens with strong coding and math at economical inference cost.
by xAI·Feb 2025
xAI's flagship reasoning LLM trained on the 100k+ GPU Colossus cluster; introduces DeepSearch and Think mode with real-time X data integration on the X platform.
by Z.ai·Sep 2025
357B MoE with 200K context (up from 128K in 4.5), MIT license; improved coding, tool-use during inference, and agentic benchmark performance.
OpenAI's second-generation reasoning model with tool use, delivering gains on math, coding, and science benchmarks over o1; released alongside o4-mini in April 2025.
by Alibaba·Nov 2024
32B dense reasoning preview model with 32K context, Apache-2.0 license; tuned for math/code chain-of-thought with strong AIME, GPQA, and MATH-500 scores.
by DeepSeek·Jan 2025
Open-weight reasoning MoE (671B total/37B active) with 128K context, MIT license; RL-trained o1-class model with strong chain-of-thought on math and code.
by OpenAI·Aug 2025
OpenAI's reasoning-native flagship succeeding the GPT-4 family, with automatic routing between fast and deep-thinking modes; released August 2025 across ChatGPT tiers.
OpenAI's compact reasoning model succeeding o3-mini, tuned for speed and cost; supports vision and tool use natively, released alongside o3 in April 2025.
by Alibaba·Apr 2025
Qwen3 open-weight family (dense 0.6B-32B + MoE 30B-A3B/235B-A22B), Apache-2.0 license; multilingual hybrid thinking modes for strong coding and math.
by Anthropic·Sep 2025
Anthropic's Claude Sonnet 4.5 mid-tier model for coding and agents; reached 77.2% on SWE-bench Verified. Priced $3/$15 per MTok.
by Google·Apr 2026
Google DeepMind embodied-reasoning model for robotics with improved spatial logic, multi-view understanding, task planning, and instrument reading.
by OpenAI·Feb 2025
OpenAI's largest pre-reasoning-era model, tuned for writing and raw capability with reduced hallucinations; served as the last GPT-4-family flagship before GPT-5.
by Anthropic·Oct 2025
Small, fast model in Anthropic's Claude 4.5 family for high-volume tasks; 200k-token context, supports extended thinking, priced $1/$5 per MTok.
by Google·Nov 2025
Google's third-generation flagship multimodal model with a 1M-token context, agent-friendly reasoning, and native support for text, image, video, and audio.
by Mistral AI·Feb 2024
Mistral's flagship closed-weights dense LLM (~123B in v2), 32k context; French-built rival to GPT-4/Claude, available via la Plateforme and Azure AI.
by Google·Mar 2025
Flagship Gemini 2.5 reasoning model with a 1M-token context window and native multimodal input across text, audio, image, video, and full code repos.
by Anthropic·Nov 2025
Anthropic's Claude Opus 4.5 for coding, agents, and computer use; introduces an API 'effort' parameter to trade off cost against capability.
Google's Apache 2.0 open-weight model family, tuned for advanced reasoning and agentic workflows and sized to run on developer hardware.
