Model Explorer

Anthropic's Mythos-class flagship with adaptive reasoning. Tops the Artificial Analysis index at launch.

Fastest Claude model. Great for high-volume, latency-sensitive applications.

Most capable Claude model. Excels at complex reasoning, agentic tasks, and extended coding.

Opus-class reasoning at a third of the old Opus price, with the context window stretched to 1M.

Near-Opus intelligence at Sonnet pricing. Strong at coding, analysis, and general tasks.

Open-source reasoning model. Competes with o1 at a fraction of the cost.

Strong general-purpose model at extremely low cost. Open weights.

Rock-bottom pricing with 1M context. Fast multimodal processing.

Cheap thinking model with 1M context. Excellent value for reasoning tasks.

Google's thinking model. Strong at coding, analysis, huge context window.

Google's streaming-first Gemini 3.5 workhorse. Powers the Live API including real-time translation.

Open-weight MoE (754B total / 40B active) built for long-horizon agentic engineering; 202K context, text-only.

Z.ai's open-weights flagship. Self-hostable frontier-adjacent capability.

Million-token context, strong at coding and instruction following.

Multimodal workhorse from OpenAI. Good all-rounder for text, vision, and audio.

Flagship OpenAI model. 400K context, strong across all tasks.

Budget GPT-5 class model with large context window.

Cheapest frontier model. On-device capable, great for high-volume simple tasks.

OpenAI frontier model unifying reasoning, coding and agentic workflows with a 1M+ token context window.

OpenAI's flagship successor to GPT-5. Instant-mode routing picks effort per request.

OpenAI's highest-compute Pro model that thinks harder for top results in agentic coding and research.

xAI's fast frontier model. Aggressive pricing for the capability tier.

Long-context reasoning model. Competitive pricing and strong performance.

Moonshot's agent-focused update to K2. Strong tool use at commodity pricing.

Compact 33B (3B active) Apache-2.0 coding model for local agentic coding; free during preview.

Strong open-source contender. 1M context, multimodal, competitive with proprietary.

10M token context. MoE architecture, self-hostable, multimodal.

Open-weight MoE (230B total / 10B active) for agentic coding; 69.4% SWE-bench Verified, text-only.

Open-weight MoE (230B total / 10B active), self-improving successor to M2; 56.2% SWE-Pro, 205K context, text-only.

European flagship. Excellent multilingual, 256K context, strong at code.

Very cheap and fast. Punches above its weight for simple to medium tasks.

Hybrid small model unifying reasoning, vision and agentic coding in one multimodal open release.

120B open hybrid Mamba-Transformer MoE (12B active) for agentic reasoning, up to 1M context.

Deep reasoning model. Excels at maths, science, coding, and logic problems.

Budget reasoning model. Good value for tasks that need structured thinking.

Top-tier Chinese open-source model. Strong at coding and multilingual tasks.

Dense 27B open-weight Qwen3.6 with built-in thinking mode, 262K context, text/image/video input.

Multimodal MoE (35B total / 3B active) Qwen3.6 with thinking control and tool calling; cheap and fast.

398B sparse MoE (~13B active) open reasoning model with explicit think traces for long-horizon agents.