#
Reasoning
AI tools tagged Reasoning.
Kimi K3
Moonshot's 2.8 trillion parameter open-weight model, the largest open model shipped to date
Ling-3.0-Flash
Ant Group's efficient open-weight reasoning model built for production AI agents
Claude Opus 4.5
Anthropic's Opus-tier model that cut pricing by two-thirds while topping SWE-bench
Claude Opus 4.6
Anthropic's Opus update that expanded the context window to 1 million tokens
Claude Opus 4.7
An Opus 4.6 upgrade targeted at the hardest software engineering tasks
Claude Opus 5
Anthropic's newest Opus-tier model, priced the same as its predecessor
Claude Sonnet 4.6
Anthropic's default Sonnet model, built to approach Opus-class results at a fifth of the price
DeepSeek V3.2
DeepSeek's V3.1 update, built around a new sparse attention mechanism
Gemini 3 Deep Think
Google's extended-reasoning mode for the hardest math and science problems
Gemini 3.1 Pro
Google's generally available successor to Gemini 3 Pro
GPT-5 Pro
The highest-compute variant of GPT-5, built for accuracy over speed
GPT-5.1
OpenAI's coding-focused GPT-5 refresh with configurable reasoning effort
GPT-5.2
OpenAI's three-tier GPT-5.2 release: Instant, Thinking, and Pro
GPT-5.4
The first OpenAI model to ship with native computer use built in
GPT-5.5
OpenAI's model that led the ARC-AGI-2 leaderboard at launch
Grok 4.20
xAI's multi-agent model that debates internally before answering
Mistral Large 4
Mistral AI's flagship reasoning and coding model
AlphaGeometry
DeepMind's neuro-symbolic system that solves olympiad geometry problems
Claude 3.7 Sonnet
Anthropic's first hybrid reasoning model, blending fast answers with extended thinking
Claude Fable 5
Anthropic's first Mythos-class model, tuned for the hardest coding benchmarks
Claude Opus 4
Anthropic's flagship model at launch, built for sustained, long-horizon agentic work
Claude Opus 4.1
An incremental refinement of Claude Opus 4 with improved coding and agentic accuracy
Claude Opus 4.8
Anthropic's most capable Opus-tier model for deep analysis and long-horizon tasks
Claude Sonnet 4
Anthropic's mid-tier Claude 4 model, balancing coding strength with everyday cost
Claude Sonnet 4.5
Anthropic's most capable Sonnet-tier model at launch, tuned heavily for coding and computer use
DeepSeek R1
DeepSeek's reasoning model that matched closed frontier models on math and code
DeepSeek V4-Pro
The top open-weight model of 2026, leading on agentic coding and graduate reasoning
DeepSeek-Math
DeepSeek's open-weight model specialized for mathematical reasoning
DeepSeek-R1-Distill-Llama-70B
A Llama-based distillation of DeepSeek-R1's reasoning traces, the largest distilled variant
DeepSeek-R1-Distill-Qwen-32B
A Qwen-based distillation of DeepSeek-R1's reasoning traces into a smaller open model
Doubao-Seed 1.6
ByteDance's reasoning-tuned tier of its Doubao foundation model
Gemini 2.0 Flash Thinking
Early experimental reasoning variant of Gemini 2.0 Flash that shows its chain of thought
Gemini 2.5 Pro
Google's 2025 reasoning-focused Gemini release
GPT-5
OpenAI's unified reasoning and chat model family
GPT-5.6 Luna
OpenAI's fast, cost-efficient GPT-5.6 tier for high-volume, latency-sensitive tasks
GPT-5.6 Sol
OpenAI's flagship model for advanced math, science, and cybersecurity reasoning
GPT-5.6 Terra
OpenAI's balanced GPT-5.6 tier for everyday coding, reasoning, and agentic tasks
GPT-OSS-120B
OpenAI's first open-weight model release since GPT-2, matching o4-mini on many reasoning benchmarks
GPT-OSS-20B
The smaller, single-GPU-friendly tier of OpenAI's open-weight model release
Grok 4
xAI's 2025 flagship reasoning model
Grok 4.5
xAI's coding-focused frontier model
Grok-1.5
xAI's transitional model that added long-context reasoning ahead of Grok 2
Hunyuan-A13B
Tencent's open-weight mixture-of-experts reasoning model
InternLM2.5
Shanghai AI Lab's updated InternLM generation with stronger tool-use and reasoning
Kimi K1.5
Moonshot's first reasoning-tuned Kimi release
Kimi K2.6
Moonshot's open-weight model built for long-running agentic and tool-use tasks
Llemma 7B
EleutherAI's open model continually pretrained on math and science text
Mathstral
Mistral's math-specialised model derived from Mistral 7B
o1
OpenAI's first reasoning model trained to think before answering
o1-mini
OpenAI's first compact reasoning model, tuned for coding and math
o1-preview
Public preview of OpenAI's first reasoning model, ahead of the full o1 release
o1-pro
Higher-compute variant of o1 for the hardest reasoning tasks
o3
OpenAI's deep multi-step reasoning model for hard technical problems
o3-mini
OpenAI's small, fast reasoning model for STEM tasks
o4-mini
OpenAI's fast, low-cost reasoning model for high-volume agentic use
Orca 2
Microsoft Research's reasoning-focused fine-tune of Llama 2
Orca-Math
Microsoft's small model specialised in grade-school math word problems
Phi-4
Microsoft's small reasoning-focused model that punches above its parameter count
Phi-4-reasoning
Reasoning-tuned variant of Phi-4 trained with chain-of-thought supervision
QwQ-32B
Alibaba's open-weight reasoning model competitive with much larger models
QwQ-32B-Preview
Alibaba's first public preview of its QwQ reasoning line, ahead of the full QwQ-32B release
Sonar Reasoning Pro
Perplexity's search-grounded reasoning model
Spark 4.0 Ultra
iFlytek's top-tier Spark model with enhanced mathematical reasoning
Step-2
StepFun's trillion-parameter reasoning model