Quick Reference: Pricing & Context
Input/output pricing per million tokens from provider docs. Notes are included where pricing tiers apply.
| Model | Pricing ($/M tokens) | Context | Best For |
|---|---|---|---|
| GPT-5.5 OpenAI | $5 / $30 | 1M | Complex reasoningCodingProfessional workflows |
| Claude Opus 4.8 Anthropic | $5 / $25 | 1M | Complex reasoningAgentic codingCritical decisions |
| GPT-5.4 OpenAI | $2.5 / $15 | 1M | CodingAgentsTool integration |
| Gemini 3.1 Pro Preview Google | $2 / $12 Standard tier, prompts <= 200k tokens; higher rates apply above 200k. | 1M | Multimodal tasksLong contextSearch integration |
| Gemini 3.5 Flash Google | $1.5 / $9 Standard tier; Batch and Flex tiers are lower. | 1M | Fast multimodal agentsSearch groundingHigh-volume workflows |
| Claude Sonnet 4.6 Anthropic | $3 / $15 | 1M | Balanced performanceProduction workloadsCost-efficient |
| GPT-5.2-Codex OpenAI | $1.75 / $14 | 400K | Coding-focused tasksType inferenceAgentic coding |
| GLM-5 Zhipu AI | $0.5 / $2 | 205K | Bilingual (CN/EN)Value-focusedEnterprise |
| Llama 4 Maverick Meta | $2 / $8 Varies by host (Together, Fireworks, etc.) | 1M | Open weightsMultimodalCustomizable |
| DeepSeek V4 Pro DeepSeek | $0.435 / $0.87 Promotional pricing listed by DeepSeek; nominal price may be higher after discounts end. | 1M | Budget codingHigh-volumeCost-sensitive |
| GPT-5.2 OpenAI | $1.75 / $14 | 400K | General-purposeBalanced tasks |
| Mistral Medium 3.5 Mistral | $2 / $6 | 128K | European complianceAgentic codingEnterprise |
| Kimi K2.5 Moonshot AI | $0.6 / $2.5 | 256K | Visual codingLong contextAgent workflows |
| MiniMax M2.5 MiniMax | $0.3 / $1.2 | 196K | Real-world productivityCost-sensitiveHigh-volume |
| Grok 4.3 xAI | $1.25 / $2.5 | 1M | Long contextWeb searchX platform data |
| Qwen3 Max 2026-01-23 Alibaba | $1.2 / $6 | 262K | MultilingualEnterpriseChinese language |
| GPT-OSS-120B OpenAI | $0 / $0 Free — open weights, self-hosted | 128K | Self-hostedPrivacyCustomization |
Performance Scores
Internal task-level evaluations across coding, reasoning, and tool-use (scale: 1-10).
| Model | Coding | Reasoning | Tool-use | Key Strengths |
|---|---|---|---|---|
| GPT-5.5 OpenAI | 9.8 | 9.8 | 9.7 |
|
| Claude Opus 4.8 Anthropic | 9.8 | 9.8 | 9.6 |
|
| GPT-5.4 OpenAI | 9.8 | 9.5 | 9.7 |
|
| Gemini 3.1 Pro Preview Google | 9.5 | 9.5 | 9.3 |
|
| Gemini 3.5 Flash Google | 9.4 | 9.4 | 9.4 |
|
| Claude Sonnet 4.6 Anthropic | 9.4 | 9.3 | 9.1 |
|
| GPT-5.2-Codex OpenAI | 9.7 | 9.3 | 9.4 |
|
| GLM-5 Zhipu AI | 9.2 | 9.3 | 9 |
|
| Llama 4 Maverick Meta | 9 | 9.1 | 8.7 |
|
| DeepSeek V4 Pro DeepSeek | 9.1 | 9.2 | 8.9 |
|
| GPT-5.2 OpenAI | 9.3 | 9.2 | 9 |
|
| Mistral Medium 3.5 Mistral | 9.1 | 9.1 | 8.9 |
|
| Kimi K2.5 Moonshot AI | 9.4 | 9.3 | 9.2 |
|
| MiniMax M2.5 MiniMax | 9.1 | 9.2 | 8.9 |
|
| Grok 4.3 xAI | 9.2 | 9.3 | 9 |
|
| Qwen3 Max 2026-01-23 Alibaba | 9.2 | 9.2 | 8.9 |
|
| GPT-OSS-120B OpenAI | 9.3 | 9.2 | 9 |
|
Quick Pick Recommendations
Not sure which model? Here's our picks by use case.
Best for Coding
Highest current coding score in the shared table, with strong reasoning and tool integration for large codebases.
Alternative: Claude Opus 4.8 for autonomous engineering reliability, or GPT-OSS-120B if you want local/self-hosted coding
Best on a Budget
Official DeepSeek pricing is unusually low for a thinking-capable 1M-context model.
Alternative: MiniMax M2.5 for broad productivity value, or Gemini 3.1 Flash-Lite for Google-native high volume
Best for Long Context
1M context plus strong reasoning makes it a safer default for large repos, large documents, and agent memory workloads.
Alternative: Claude Opus 4.8 or Gemini 3.5 Flash when provider fit matters more than OpenAI ecosystem
Best for Reasoning
Strong reasoning plus practical tool-use makes it one of the most useful frontier picks right now.
Alternative: Qwen3 Max 2026-01-23 for bilingual enterprise reasoning
Best All-Rounder
Best blend of coding, reasoning, tool use, context, and first-party platform support on the board.
Alternative: Claude Opus 4.8 if reliability and professional work quality matter more than OpenAI tooling
Best for Local / Open Weights
Best open-weights option here if you care about privacy, self-hosting, and no per-token API bills.
Alternative: Llama 4 or Qwen3-Coder for lighter local deployments
Need a Faster Decision?
Most competitors win on discoverability because they guide the user to the next step immediately. This does that.
Choose an API model
Use this if you care about fastest setup, no infrastructure, and provider-managed uptime.
Browse ranked API models →Run models locally
Use open-weights models if privacy, customization, or long-term cost control matters most.
See local model guide →Understand open vs proprietary
Still not sure? Start with the tradeoffs around privacy, cost, customization, and performance.
Read the comparison →Verification Sources
Official model and pricing references checked on 2026-06-03.
| Model | Official Sources |
|---|---|
| GPT-5.5 OpenAI | |
| Claude Opus 4.8 Anthropic | |
| GPT-5.4 OpenAI | |
| Gemini 3.1 Pro Preview Google | |
| Gemini 3.5 Flash Google | |
| Claude Sonnet 4.6 Anthropic | |
| GPT-5.2-Codex OpenAI | |
| GLM-5 Zhipu AI | |
| Llama 4 Maverick Meta | |
| DeepSeek V4 Pro DeepSeek | |
| GPT-5.2 OpenAI | |
| Mistral Medium 3.5 Mistral | |
| Kimi K2.5 Moonshot AI | |
| MiniMax M2.5 MiniMax | |
| Grok 4.3 xAI | |
| Qwen3 Max 2026-01-23 Alibaba | |
| GPT-OSS-120B OpenAI |
Dive Deeper into Model Performance
See detailed daily scorecards with task-level breakdowns, failure cases, and cost analysis.
View Daily Scorecards → Our Methodology