Compare the top AI models for code generation, debugging, and software development. Real pricing, context windows, and capabilities from first-party data.
The most capable models for complex coding tasks. Higher price, highest quality.
| Model | Provider | Input $/1M | Output $/1M | Context | Tool Call | Reasoning |
|---|---|---|---|---|---|---|
| gpt-4.1 | openai | $2 | $8 | 1M | ✅ | |
| gpt-4o | openai | $2.5 | $10 | 128K | ✅ | |
| gemini-2.5-pro | deepinfra | $1.25 | $10 | 1M | ✅ | |
| deepseek-r1 | amazon-bedrock | $1.35 | $5.4 | 65K |
Great coding performance at lower prices. Perfect for high-volume code generation.
| Model | Provider | Input $/1M | Output $/1M | Context | Tool Call | Reasoning |
|---|---|---|---|---|---|---|
| gpt-4o-mini | openai | $0.15 | $0.6 | 128K | ✅ | |
| gemini-2.5-flash | deepinfra | $0.3 | $2.5 | 1M | ✅ | |
| deepseek-v3 | deepinfra | $0.32 | $0.89 | 163K | ||
| deepseek-r1 | amazon-bedrock | $1.35 | $5.4 | 65K |
Zero-cost models for learning, prototyping, and personal projects.
| Model | Provider | Context | Tool Call | Reasoning |
|---|---|---|---|---|
| openrouter--owl-alpha | openrouter | 1M | ✅ | |
| deepseek--deepseek-v4-flash--free | openrouter | 1M | ✅ | ✅ |
| qwen--qwen3-coder--free | openrouter | 1M | ✅ | |
| nvidia--nemotron-3-super-120b-a12b--free | openrouter | 1M | ✅ | ✅ |
| gemma-4-26b-a4b-it | auriko | 262K | ✅ | ✅ |
| gemma-4-31b-it | auriko | 262K | ✅ | ✅ |
| arcee-ai--trinity-large-thinking--free | openrouter | 262K | ✅ | ✅ |
| google--gemma-4-26b-a4b-it--free | openrouter | 262K | ✅ | ✅ |
| google--gemma-4-31b-it--free | openrouter | 262K | ✅ | ✅ |
| nvidia--nemotron-3-nano-omni-30b-a3b-reasoning--free | openrouter | 256K | ✅ | ✅ |
Download and run locally for full privacy and zero API costs at scale.
| Model | Provider | Context | Tool Call | Reasoning |
|---|---|---|---|---|
| google--gemma-4-31b-it | orcarouter | 1M | ✅ | |
| qwen--qwen3.5-flash-2026-02-23 | orcarouter | 1M | ✅ | |
| qwen--qwen3.5-flash | orcarouter | 1M | ✅ | |
| qwen--qwen3.6-flash-2026-04-16 | orcarouter | 1M | ✅ | |
| qwen--qwen3.6-flash | orcarouter | 1M | ✅ | |
| meta-llama-4-maverick-17b | amazon-bedrock | 1M | ✅ | |
| meta-llama-4-scout-17b | amazon-bedrock | 1M | ✅ | |
| minimax-m2-1 | amazon-bedrock | 1M | ✅ | |
| minimax-m2-5 | amazon-bedrock | 1M | ✅ | |
| minimax-m2 | amazon-bedrock | 1M | ✅ |
Models with 128K+ context for working with large codebases, multiple files, and long conversations.
| Model | Provider | Context | Input $/1M | Tool Call |
|---|---|---|---|---|
| ling-2.6-flash | inclusionai | 262K | $0.01 | ✅ |
| bdc-coder | inferencenet | 131K | $0.01 | ✅ |
| klusterai--Meta-Llama-3.1-8B-Instruct-Turbo | klusterai | 131K | $0.015 | ✅ |
| granite-4.0-h-micro | cloudflare | 131K | $0.017 | ✅ |
| llama-3.1-8b-instruct--fp-16 | inferencenet | 131K | $0.02 | ✅ |
| schematron-3b | inferencenet | 131K | $0.02 | ✅ |
| schematron-v3 | inferencenet | 131K | $0.02 | ✅ |
| gpt-oss-20b | inferencenet | 131K | $0.03 | ✅ |
| schematron-v2-turbo | inferencenet | 131K | $0.03 | ✅ |
| qwen--qwen3-4b-fp8 | novitaai | 128K | $0.03 | ✅ |
| liquid-ai--LFM2-24B-A2B | togetherai | 131K | $0.03 | ✅ |
| amazon-nova-micro | amazon | 128K | $0.035 | ✅ |
| amazon-nova-micro | amazon-bedrock | 128K | $0.035 | ✅ |
| mistral-nemo-12b-instruct--fp-8 | inferencenet | 131K | $0.0375 | ✅ |
| klusterai--Meta-Llama-3.3-70B-Instruct-Turbo | klusterai | 131K | $0.038 | ✅ |
Models with tool calling + reasoning — the key capabilities for AI coding agents (Cursor, Copilot, Devin-style).
| Model | Provider | Input $/1M | Output $/1M | Context |
|---|---|---|---|---|
| openai--gpt-oss-20b | neuralwatt | $0.03 | $0.16 | ? |
| qwen--qwen3-4b-fp8 | novitaai | $0.03 | $0.03 | 128K |
| gpt-oss-120b | inferencenet | $0.05 | $0.45 | 131K |
| Qwen--Qwen3.6-35B-A3B | neuralwatt | $0.05 | $0.1 | ? |
| openai--gpt-oss-120b | novitaai | $0.05 | $0.25 | 131K |
| qwen3-30b-a3b-fp8 | cloudflare | $0.051 | $0.335 | 40K |
| glm-4.7-flash | cloudflare | $0.06 | $0.4 | 131K |
| Nemotron-3-Nano-Omni | nebius | $0.06 | $0.24 | 128K |
| hermes-4-llama-3.1-8b | nousresearch | $0.06 | $0.12 | 131K |
| seed-1.6-flash | bytedance | $0.07 | $0.3 | 262K |
| ring-2.6-1t | inclusionai | $0.07 | $0.62 | 262K |
| zai-org--glm-4.7-flash | novitaai | $0.07 | $0.4 | 200K |
| microsoft-phi-4-mini-reasoning | microsoft | $0.075 | $0.3 | 128K |
| Qwen--Qwen3-32B-TEE | chutes | $0.08 | $0.24 | 40K |
| gpt-oss-120b | clarifai | $0.09 | $0.36 | 131K |
All data is sourced from first-party APIs. Models are selected based on capabilities relevant to coding: tool calling (for agentic workflows), reasoning (for complex logic), large context (for codebases), and structured output (for parsing). Aggregator providers are excluded from ranking tables.