Code

Open-weight 117B MoE model (5.1B active) achieving near o4-mini reasoning. Apache 2.0 licensed, runs on a single 80GB GPU.

Poolside's Laguna XS 2.1 — compact coding-agent model with a 262K-token context window.

Qwen3 Coder 30B A3B Instruct, an open-weight Alibaba model available via OpenRouter.

Poolside's Laguna S 2.1 coding-agent model with a 1M-token context window.

Qwen3 Coder Next, an open-weight Alibaba model available via OpenRouter.

Reasoning-focused model with chain-of-thought capabilities rivaling o1.

Latest DeepSeek V3 with improved reasoning and coding. 671B MoE (37B active), MIT licensed, 164K context.

Qwen3 Coder Flash, an open-weight Alibaba model available via OpenRouter.

Open-weight 671B MoE model with strong coding and reasoning at low cost.

Updated DeepSeek V3 with improved coding and reasoning performance.

GPT-5.1-Codex-Mini, a proprietary OpenAI model available via OpenRouter.

Qwen3 Coder 480B A35B, an open-weight Alibaba model available via OpenRouter.

Open-weight multimodal model with agent swarm mode supporting up to 100 parallel sub-agents.

Updated R1 with improved reasoning accuracy and reduced hallucination.

GPT-5 variant optimized for code generation and software engineering.

GPT-5.1 code-optimized variant.

State-of-the-art 1T MoE model with 32B active parameters. Strong coding and agentic capabilities.

ByteDance Seed 2.0 Code — coding-specialized Seed 2.0 model with image/video input and a 262K-token context window.

Frontier 744B model trained on Huawei Ascend chips. Open source with strong agentic capabilities.

Qwen3 Coder Plus, an open-weight Alibaba model available via OpenRouter.

Coding-focused agentic model built on Kimi K2.6 with ~30% lower thinking-token usage. 1T MoE (32B active), native INT4 quantization, 256K context, text/image/video input. Forces thinking and preserve_thinking modes; tuned for long-horizon software engineering and agent workflows.

Moonshot AI's open-weight flagship: natively multimodal 1T MoE model (32B active) with text, image, and video input, thinking and non-thinking modes, and 256K context. State-of-the-art open-source coding and agentic performance (SWE-Bench Verified 80.2); Agent Swarm scales to 300 parallel sub-agents.

GPT-5.2 code-optimized variant.

GPT-5.3 code-optimized variant.

OpenAI's latest frontier model combining reasoning, coding, and agentic workflows.

Latest GPT-5.4 code-optimized variant with industry-leading coding capabilities.

Kwaipilot's KAT-Coder-Pro V2.5 agentic coding model with a 262K-token context window.

Latest Sonnet with Opus-tier capabilities at Sonnet pricing.

xAI's Grok Build 0.1 coding-agent model.

Coding-focused frontier model scoring 94% of Claude Opus 4.6. 744B MoE trained on Huawei Ascend 910B. #1 on SWE-Bench Pro (open source).

GPT-5.1-Codex-Max, a proprietary OpenAI model available via OpenRouter.

Anthropic's most capable model with 1M token context and advanced reasoning.

Reasoning-focused model with step-by-step deliberation for complex math, coding, and science tasks.

The high-tier variant of Sakana AI's Fugu orchestration model. Reportedly outperforms Opus 4.8, GPT-5.5 and Gemini 3.1 Pro on many benchmarks (73.7 on SWE-bench Pro), standing shoulder-to-shoulder with Claude Fable 5. Built on the TRINITY and Conductor (ICLR 2026) research.

Most capable reasoning model in OpenAI's lineup with extended thinking for maximum reliability.

Highest capability GPT-5.4 tier with maximum reasoning depth. Premium pricing.