OpenCode Models Guide

August 22, 2026 · View on GitHub

Comprehensive guide to OpenCode Go and Zen models with capabilities, costs, and routing recommendations.

Sources: OpenCode Go Documentation | OpenCode Zen Documentation

Quick Cost Comparison

💰 Cost-conscious routing matters! Qwen3.5 Plus gives you 10,200 requests per $12, while GLM-5.1 gives you only 880 — that's 11.6x fewer requests for the same budget.

Note: the requests-per-$12 figures below are approximate estimates for comparison only — they are not derived from a machine-readable price list, and the model catalog carries no rate data. Treat the ordering as meaningful and the absolute numbers as indicative. Check your provider's current pricing before budgeting.

ModelProviderRequests per $12 (5hr)Cost EfficiencyQuality
Qwen3.5 PlusGo10,200★★★★★★★☆☆☆
MiniMax M2.5Go6,300★★★★★★★☆☆☆
Qwen3.7 PlusGo4,300★★★★★★★★☆☆
MiniMax M2.7Go3,400★★★★☆★★★☆☆
MiniMax M3Go3,200★★★★☆★★★☆☆
Qwen3.6 PlusGo3,300★★★★☆★★★☆☆
MiMo-V2.5Go2,150★★★☆☆★★★☆☆
MiMo-V2.5-ProGo1,290★★☆☆☆★★★★☆
Kimi K2.5Go1,850★★☆☆☆★★★★☆
Kimi K2.6Go1,850★★☆☆☆★★★★★
Kimi K2.7 CodeGo1,350★☆☆☆☆★★★★★
Kimi K3Go$3/$15 per 1M☆☆☆☆☆★★★★★
GLM-5Go1,150★☆☆☆☆★★★★☆
GLM-5.1Go880☆☆☆☆☆★★★★★
GLM-5.2Go880☆☆☆☆☆★★★★★
Qwen3.7 MaxGo950☆☆☆☆☆★★★★☆

Providers

OpenCode Go (opencode-go)

  • Subscription-based ($5/month then $10/month)
  • OpenAI Chat Completions and Anthropic Messages endpoints
  • Best for: Most use cases, cost-effective models

OpenCode Zen (opencode-zen)

  • Pay-as-you-go pricing
  • Additional endpoint formats: Responses (GPT), Gemini
  • Best for: GPT models, Gemini models, premium Anthropic models

AWS Bedrock (aws-bedrock)

  • Models hosted on AWS Bedrock Mantle
  • Supports OpenAI Chat Completions (default) and Anthropic Messages formats
  • Set wire_format: "anthropic" for Claude and other Anthropic-native models
  • Best for: Models deployed on your own AWS infrastructure

OpenRouter Models

OpenRouter provides unified access to 100+ models from multiple providers through a single API endpoint. Instead of managing separate API keys and endpoints for each provider, you can access models from OpenAI, Anthropic, Google, Meta, Mistral, and many others through one integration.

Key Benefits

BenefitDescription
Unified APISingle OpenAI-compatible Chat Completions endpoint for all models
100+ ModelsAccess to models from 20+ providers without separate integrations
Dynamic CatalogNew models automatically available when the catalog is updated
Pay-as-you-goPer-token pricing with no subscription required
Smart RoutingAutomatic fallback to alternative providers if a model is unavailable
Cost OptimizationRoute to the cheapest or fastest available provider

Model Naming Convention

OpenRouter uses a provider/model-name format that identifies both the original provider and the specific model:

FormatExampleDescription
openai/gpt-4oopenai/gpt-4oGPT-4o via OpenAI
openai/gpt-4o-miniopenai/gpt-4o-miniGPT-4o Mini via OpenAI
anthropic/claude-3.5-sonnetanthropic/claude-3.5-sonnetClaude 3.5 Sonnet via Anthropic
anthropic/claude-3-opusanthropic/claude-3-opusClaude 3 Opus via Anthropic
google/gemini-2.5-progoogle/gemini-2.5-proGemini 2.5 Pro via Google
google/gemini-2.0-flashgoogle/gemini-2.0-flashGemini 2.0 Flash via Google
meta-llama/llama-3.3-70bmeta-llama/llama-3.3-70bLlama 3.3 70B via Meta
mistral/mistral-largemistral/mistral-largeMistral Large via Mistral AI
x-ai/grok-2x-ai/grok-2Grok 2 via xAI
deepseek/deepseek-chatdeepseek/deepseek-chatDeepSeek V3 via DeepSeek
ModelProviderContext WindowInput Cost ($/M)Output Cost ($/M)Best For
Claude 3.5 SonnetAnthropic200K$3.00$15.00Complex reasoning, coding, analysis
Claude 3 OpusAnthropic200K$15.00$75.00Maximum quality, difficult tasks
GPT-4oOpenAI128K$2.50$10.00General purpose, vision tasks
GPT-4o MiniOpenAI128K$0.15$0.60Cost-effective, high volume
Gemini 2.5 ProGoogle1M$1.25$10.00Long context, coding, reasoning
Gemini 2.0 FlashGoogle1M$0.10$0.40Fast responses, cost efficiency
Llama 3.3 70BMeta128K$0.12$0.30Open source, customizable
Llama 3.1 405BMeta128K$0.80$1.60Open source, high quality
Mistral LargeMistral128K$2.00$6.00European provider, GDPR compliant
Grok 2xAI128K$2.00$10.00Real-time knowledge, humor
DeepSeek V3DeepSeek64K$0.07$1.10Cost efficiency, coding
DeepSeek R1DeepSeek64K$0.55$2.19Reasoning, chain-of-thought

Discovering Available Models

Browse the complete model catalog at: https://openrouter.ai/models

The catalog includes detailed information for each model:

  • Pricing: Input and output costs per million tokens
  • Context window: Maximum tokens the model can process
  • Tool calling: Whether the model supports function calling
  • Vision: Whether the model supports image input
  • Provider routing: Available providers for each model

You can also fetch available models programmatically via the OpenRouter API:

curl https://openrouter.ai/api/v1/models \
  -H "Authorization: Bearer $OPENROUTER_API_KEY"

Configuring OpenRouter Models

Add OpenRouter models to your routatic-proxy configuration:

{
  "models": {
    "default": {
      "provider": "openrouter",
      "model_id": "anthropic/claude-3.5-sonnet",
      "temperature": 0.7,
      "max_tokens": 4096
    },
    "background": {
      "provider": "openrouter",
      "model_id": "openai/gpt-4o-mini",
      "temperature": 0.5,
      "max_tokens": 2048
    },
    "complex": {
      "provider": "openrouter",
      "model_id": "anthropic/claude-3-opus",
      "temperature": 0.7,
      "max_tokens": 8192
    },
    "long_context": {
      "provider": "openrouter",
      "model_id": "google/gemini-2.5-pro",
      "temperature": 0.7,
      "max_tokens": 8192
    }
  }
}

Cost-Based Routing Integration

When cost_routing.enabled is set, the selector automatically sorts models by cost and applies your routing preferences:

{
  "cost_routing": {
    "enabled": true,
    "prefer_providers": ["opencode-go", "openrouter"],
    "penalty_per_provider": {
      "openrouter": 0.05
    },
    "max_context_window": 200000
  }
}

How cost routing works with OpenRouter:

  • Models are sorted by combined input + output rates
  • Provider penalties adjust effective costs (e.g., openrouter: 0.05 adds 5%)
  • max_context_window filters out models that can't handle your request size
  • prefer_providers intersects with scenario preferences for final selection

Model Categories Available

CategoryProvidersExample Models
OpenAIOpenAI, AzureGPT-4o, GPT-4o Mini, GPT-4 Turbo
AnthropicAnthropic, AWSClaude 3 Opus, Claude 3.5 Sonnet, Claude 3 Haiku
GoogleGoogleGemini 2.5 Pro, Gemini 2.0 Flash, Gemini 1.5 Pro
MetaMeta, Together, FireworksLlama 3.3 70B, Llama 3.1 405B, Llama 3.1 70B
MistralMistral AIMistral Large, Mistral Medium, Mistral Small
xAIxAIGrok 2, Grok Beta
DeepSeekDeepSeekDeepSeek V3, DeepSeek R1
SpecializedVariousQwen, Command R+, Perplexity, and many more

Example Configurations

Budget-Conscious Setup

Use cheaper models for most tasks, expensive ones only when needed:

{
  "models": {
    "background": {
      "provider": "openrouter",
      "model_id": "openai/gpt-4o-mini"
    },
    "default": {
      "provider": "openrouter",
      "model_id": "deepseek/deepseek-chat"
    },
    "complex": {
      "provider": "openrouter",
      "model_id": "anthropic/claude-3.5-sonnet"
    }
  }
}

Quality-First Setup

Prioritize quality for critical tasks:

{
  "models": {
    "default": {
      "provider": "openrouter",
      "model_id": "anthropic/claude-3.5-sonnet"
    },
    "complex": {
      "provider": "openrouter",
      "model_id": "anthropic/claude-3-opus"
    },
    "long_context": {
      "provider": "openrouter",
      "model_id": "google/gemini-2.5-pro"
    }
  }
}

Multi-Provider Fallback

Spread requests across providers for reliability:

{
  "models": {
    "default": {
      "provider": "openrouter",
      "model_id": "openai/gpt-4o"
    },
    "fallback": {
      "provider": "openrouter",
      "model_id": "anthropic/claude-3.5-sonnet"
    }
  },
  "fallbacks": {
    "default": [
      { "model_id": "openai/gpt-4o" },
      { "model_id": "anthropic/claude-3.5-sonnet" },
      { "model_id": "google/gemini-2.5-pro" }
    ]
  }
}

OpenRouter-Specific Features

Provider Routing Preferences

Request specific providers or enable automatic fallback:

{
  "model_overrides": {
    "claude-sonnet": {
      "provider": "openrouter",
      "model_id": "anthropic/claude-3.5-sonnet",
      "temperature": 0.7,
      "max_tokens": 4096,
      "extra_body": {
        "provider": {
          "order": ["Anthropic", "AWS"],
          "allow_fallbacks": true
        }
      }
    }
  }
}

Routing options:

  • order: Priority list of providers to try
  • allow_fallbacks: Whether to try other providers if primary is down
  • ignore: Providers to exclude from routing

How OpenRouter Model Catalog Works

OpenRouter models are resolved dynamically from the catalog system:

  1. Catalog Location: ~/.config/routatic-proxy/catalog/catalog.json
  2. Resolution: Models are keyed as provider/model-name (e.g., openai/gpt-4o, anthropic/claude-3.5-sonnet)
  3. Dynamic Loading: New models are automatically available when the catalog is updated — no code changes required

Model Resolution from Catalog

The catalog system extracts the provider from the key prefix:

{
  "providers": {
    "openrouter": {
      "name": "OpenRouter",
      "base_url": "https://openrouter.ai/api/v1",
      "enabled": true
    }
  },
  "models": {
    "openrouter/anthropic/claude-3.5-sonnet": {
      "id": "openrouter/anthropic/claude-3.5-sonnet",
      "name": "Claude 3.5 Sonnet",
      "limit": { "context": 200000 },
      "rates": { "input": 3.0, "output": 15.0 },
      "tool_call": true,
      "modalities": { "input": ["text", "image"], "output": ["text"] },
      "reasoning": false
    }
  }
}
  • ResolvedModel.ModelID: The model name without provider prefix (anthropic/claude-3.5-sonnet)
  • ResolvedModel.CanonicalName: The full key (openrouter/anthropic/claude-3.5-sonnet)

Benefits Summary

  1. Access to 100+ Models: Single API key for OpenAI, Anthropic, Google, Meta, Mistral, and more
  2. Unified API: OpenAI-compatible Chat Completions format for all models
  3. Automatic Fallbacks: Built-in fallback to alternative providers if a model is unavailable
  4. Cost Optimization: Per-token pricing with routing preferences for cost efficiency
  5. No Vendor Lock-in: Switch between providers by changing the model ID

Important: API Endpoints

⚠️ Critical: Not all models use the same API endpoint! routatic-proxy handles this automatically, but you should know:

OpenCode Go Endpoints

ModelsEndpointFormat
GLM-5, GLM-5.1, GLM-5.2, Kimi K2.5, Kimi K2.6, Kimi K2.7 Code, Kimi K3, MiMo-V2.5, MiMo-V2.5-Pro, DeepSeek V4 Pro, DeepSeek V4 Flashhttps://opencode.ai/zen/go/v1/chat/completionsOpenAI-compatible
MiniMax M2.5, MiniMax M2.7, MiniMax M3, Qwen3.5 Plus, Qwen3.6 Plus, Qwen3.7 Plus, Qwen3.7 Maxhttps://opencode.ai/zen/go/v1/messagesAnthropic-compatible

OpenCode Zen Endpoints

ModelsEndpointFormat
MiniMax M2.5, MiniMax M2.7, MiniMax M3, GLM-5, GLM-5.1, GLM-5.2, Kimi K2.5, Kimi K2.6, Kimi K2.7 Code, Kimi K3, DeepSeek V4 Pro, DeepSeek V4 Flash, DeepSeek V4 Flash Free, Grok Build 0.1, Big Pickle, MiMo-V2.5 Free, North Mini Code Free, Nemotron 3 Ultra Freehttps://opencode.ai/zen/v1/chat/completionsOpenAI-compatible
Claude models (claude-fable-5, claude-opus-4-8, claude-opus-4-7, claude-opus-4-6, claude-opus-4-5, claude-opus-4-1, claude-sonnet-4-6, claude-sonnet-4-5, claude-sonnet-4, claude-haiku-4-5, claude-3-5-haiku), Qwen models (qwen3.5-plus, qwen3.6-plus, qwen3.7-plus, qwen3.7-max)https://opencode.ai/zen/v1/messagesAnthropic-compatible
GPT models (gpt-5.5, gpt-5.5-pro, gpt-5.5-mini, gpt-5.5-nano, gpt-5.4, gpt-5.4-pro, gpt-5.4-mini, gpt-5.4-nano, gpt-5.3-codex, gpt-5.3-codex-spark, gpt-5.2, gpt-5.2-codex, gpt-5.1, gpt-5.1-codex, gpt-5.1-codex-max, gpt-5.1-codex-mini, gpt-5, gpt-5-codex, gpt-5-nano)https://opencode.ai/zen/v1/responsesOpenAI Responses
Gemini models (gemini-3.5-flash, gemini-3.1-pro, gemini-3-flash)https://opencode.ai/zen/v1/models/{id}Google Gemini

Why this matters: On the Go provider, MiniMax and Qwen models use Anthropic format natively. On Zen, only Claude and Qwen use the Anthropic endpoint — MiniMax uses chat completions. routatic-proxy handles all routing automatically.

Using OpenCode Zen

To use Zen models, set "provider": "opencode-zen" in your model config:

{
  "models": {
    "default": {
      "provider": "opencode-zen",
      "model_id": "kimi-k2.6",
      "temperature": 0.7,
      "max_tokens": 4096
    }
  }
}

Zen-Specific Models (50+ total)

All OpenCode Go models are also available on Zen. Zen additionally offers:

  • Claude Models (Anthropic endpoint): claude-fable-5, claude-opus-4-8, claude-opus-4-7, claude-opus-4-6, claude-opus-4-5, claude-opus-4-1, claude-sonnet-4-6, claude-sonnet-4-5, claude-sonnet-4, claude-haiku-4-5, claude-3-5-haiku
  • GPT Models (Responses endpoint): gpt-5.5, gpt-5.5-pro, gpt-5.5-mini, gpt-5.5-nano, gpt-5.4, gpt-5.4-pro, gpt-5.4-mini, gpt-5.4-nano, gpt-5.3-codex, gpt-5.3-codex-spark, gpt-5.2, gpt-5.2-codex, gpt-5.1, gpt-5.1-codex, gpt-5.1-codex-max, gpt-5.1-codex-mini, gpt-5, gpt-5-codex, gpt-5-nano
  • Gemini Models (Gemini endpoint): gemini-3.5-flash, gemini-3.1-pro, gemini-3-flash
  • Free Tier (chat completions): deepseek-v4-flash-free, big-pickle, mimo-v2.5-free, north-mini-code-free, nemotron-3-ultra-free

Deprecated Zen Models

The following models are deprecated and will be removed:

ModelDeprecation DateReplacement
GPT 5.2 CodexJuly 23, 2026GPT 5.3 Codex
GPT 5.1 CodexJuly 23, 2026GPT 5.3 Codex
GPT 5.1 Codex MaxJuly 23, 2026GPT 5.3 Codex
GPT 5.1 Codex MiniJuly 23, 2026GPT 5.3 Codex Spark
GPT 5 CodexJuly 23, 2026GPT 5.3 Codex
Claude Sonnet 4June 15, 2026Claude Sonnet 4.5/4.6
GLM 5May 14, 2026GLM 5.1/5.2
MiniMax M2.1March 15, 2026MiniMax M2.5/M2.7/M3
GLM 4.7March 15, 2026GLM 5/5.1/5.2
GLM 4.6March 15, 2026GLM 5/5.1/5.2
Gemini 3 ProMarch 9, 2026Gemini 3.1 Pro
Kimi K2 ThinkingMarch 6, 2026Kimi K2.5/K2.6/K2.7 Code
Kimi K2March 6, 2026Kimi K2.5/K2.6/K2.7 Code
Claude Haiku 3.5Feb 16, 2026Claude Haiku 4.5
Qwen3 Coder 480BFeb 6, 2026Qwen3.7 Plus/Max

DeepSeek V4 Pro and Flash are OpenAI-compatible on both Go and Zen providers. DeepSeek V4 Flash Free is the free Zen variant. routatic-proxy transforms Claude Code's Anthropic request into OpenAI Chat Completions format, including tools, tool results, thinking history, reasoning_effort, and thinking.

For Claude Code and OpenCode-style agent workflows, DeepSeek V4 supports max thinking mode with:

{
  "model_id": "deepseek-v4-pro",
  "reasoning_effort": "max",
  "thinking": {
    "type": "enabled"
  }
}

Use deepseek-v4-pro for default, complex, thinking, and long-context routing. Use deepseek-v4-flash for fast, background, or subagent-style workloads.

To route DeepSeek V4 Pro through Zen (free tier) instead of Go (paid), add a model_overrides entry:

{
  "model_overrides": {
    "deepseek-v4-pro": {
      "provider": "opencode-zen",
      "model_id": "deepseek-v4-pro",
      "temperature": 0.7,
      "max_tokens": 8192,
      "reasoning_effort": "max",
      "thinking": {
        "type": "enabled"
      }
    }
  }
}

Cost-Conscious Routing Strategy

Default to Cheap, Upgrade When Necessary

Most requests should use cheap models. Only upgrade to expensive models when:

  1. Task complexity demands it (multi-step reasoning, architecture)
  2. You've tried cheaper models and they failed
  3. Code quality is critical (production code review)
{
  "models": {
    "background": {
      // Simple operations
      "model_id": "qwen3.5-plus",
      "max_tokens": 2048
    },
    "default": {
      // Better quality, moderate cost
      "model_id": "kimi-k2.6",
      "max_tokens": 4096
    },
    "long_context": {
      // Large files only — needs a 1M-context model
      "model_id": "minimax-m3",
      "context_threshold": 100000
    },
    "think": {
      // Reasoning tasks
      "model_id": "glm-5",
      "max_tokens": 8192
    },
    "complex": {
      // Complex architecture only
      "model_id": "glm-5.1",
      "max_tokens": 4096
    },
    "fast": {
      // Streaming requests (prioritize TTFT)
      "model_id": "qwen3.6-plus",
      "max_tokens": 4096
    }
  }
}

Decision Tree

Is context > 100K tokens? (default threshold, configurable via context_threshold)
├── YES → Use MiniMax M3 (1M context, 3,200 req/\$12)
│
Is it a complex task (architecture, refactoring, tool operations)?
├── YES → Use GLM-5.1 (880 req/\$12)
│
Is it a reasoning/planning task?
├── YES → Use GLM-5 (1,150 req/\$12)
│
Is it a simple background task (read file, grep, list dir, no tools)?
├── YES → Use Qwen3.5 Plus (10,200 req/\$12)
│
Default → Use Kimi K2.6 (1,850 req/\$12, ★★★★★) or Qwen3.6 Plus (3,300 req/\$12)

Detailed Model Profiles

Budget Champions 💰

Qwen3.5 Plus — The Workhorse

  • Model ID: qwen3.5-plus
  • Cost: 10,200 requests per $12 (best value!)
  • Context: ~1M tokens
  • Quality: ★★☆☆☆ (adequate for simple tasks)
  • Modalities: Text and image input
  • Best For:
    • File reading operations
    • Directory listing
    • Grep/search
    • Simple questions
    • Bulk operations
    • Background tasks
  • When to Use: When you need to do lots of operations cheaply

MiniMax M2.5 — Cheapest 200K-Class Model

  • Model ID: minimax-m2.5
  • Endpoint: Anthropic-compatible (/v1/messages on Go), OpenAI-compatible (/chat/completions on Zen)
  • Cost: 6,300 requests per $12
  • Context: ~200K tokens
  • Max Output: 4K tokens
  • Quality: ★★☆☆☆ (acceptable)
  • Speed: Fast
  • Best For:
    • Large files that still fit inside 200K
    • Long conversations on a tight budget
    • Multi-file context
  • When to Use: When 200K of context is enough and cost is the priority. For genuinely long context (>100K, up to 1M) use MiniMax M3 instead.
  • Note: Uses Anthropic endpoint on Go but chat completions on Zen - routatic-proxy handles this automatically

MiniMax M3 — Latest MiniMax, 1M Context

  • Model ID: minimax-m3
  • Endpoint: Anthropic-compatible (/v1/messages on Go), OpenAI-compatible (/chat/completions on Zen)
  • Cost: 3,200 requests per $12
  • Context: ~1M tokens
  • Max Output: 128K tokens
  • Quality: ★★★☆☆
  • Best For:
    • Long-context tasks (the recommended long_context model)
    • Large codebase analysis
    • Document processing
  • When to Use: Whenever the request exceeds the long-context threshold — M2.5 tops out at 200K, M3 goes to 1M

Balanced Models (Quality + Cost)

DeepSeek V4 Pro — Agentic Coding + Max Thinking

  • Model ID: deepseek-v4-pro

  • Endpoint: OpenAI-compatible (/chat/completions)

  • Context: ~1M tokens

  • Quality: ★★★★★

  • Providers: Go (paid) or Zen (free tier)

  • Best For:

    • Claude Code agent workflows
    • Complex implementation and debugging
    • Architecture and refactoring
    • Long-context coding tasks
    • Max thinking mode

  • Recommended Config (Go):

    {
  "provider": "opencode-go",
  "model_id": "deepseek-v4-pro",
  "temperature": 0.1,
  "max_tokens": 8192,
  "reasoning_effort": "max",
  "thinking": {
    "type": "enabled"
  }
}

  • Recommended Config (Zen free tier):

    {
  "provider": "opencode-zen",
  "model_id": "deepseek-v4-pro",
  "temperature": 0.1,
  "max_tokens": 8192,
  "reasoning_effort": "max",
  "thinking": {
    "type": "enabled"
  }
}

DeepSeek V4 Flash — Fast Agent Workloads

  • Model ID: deepseek-v4-flash

  • Endpoint: OpenAI-compatible (/chat/completions)

  • Context: ~1M tokens

  • Quality: ★★★★☆

  • Best For:

    • Fast routing
    • Background tasks
    • Subagent-style work
    • Fallback for DeepSeek V4 Pro

  • Recommended Config:

    {
  "provider": "opencode-go",
  "model_id": "deepseek-v4-flash",
  "temperature": 0.1,
  "max_tokens": 4096,
  "reasoning_effort": "max",
  "thinking": {
    "type": "enabled"
  }
}
  • Model ID: qwen3.6-plus
  • Endpoint: Anthropic-compatible (/v1/messages — Go), Anthropic-compatible (/v1/messages — Zen)
  • Cost: 3,300 requests per $12 (3.8x more than GLM-5.1!)
  • Context: ~1M tokens
  • Quality: ★★★☆☆ (good enough for most tasks)
  • Modalities: Text and image input
  • Speed: Fast
  • Best For:
    • General coding (default choice)
    • Feature implementation
    • Bug fixes
    • Refactoring
  • When to Use: Default for cost-conscious users

Qwen3.7 Plus / Max — see the Premium Models section below.

Kimi K2.6 — Best Quality at Balanced Cost

  • Model ID: kimi-k2.6
  • Cost: ~1,850 requests per $12
  • Context: ~256K tokens (successor to K2.5 with improvements)
  • Quality: ★★★★★ (excellent — successor improvements)
  • Modalities: Text and image input
  • Speed: Fast
  • Best For:
    • Complex coding tasks
    • Code review
    • Architecture discussions
    • General-purpose default (best quality-to-cost ratio)
  • When to Use: Default choice — better quality than K2.5 at similar cost

Kimi K2.5 — Quality + Reasonable Cost (Predecessor)

  • Model ID: kimi-k2.5
  • Cost: 1,850 requests per $12
  • Context: ~256K tokens (2x most others)
  • Quality: ★★★★☆ (excellent)
  • Modalities: Text and image input
  • Speed: Fast
  • Best For:
    • Complex coding tasks
    • Code review
    • Architecture discussions
    • When you need better quality than budget models
  • When to Use: When quality matters more than maximum cost savings

Premium Models (Use Sparingly!)

GLM-5 — Reasoning Specialist

  • Model ID: glm-5
  • Cost: 1,150 requests per $12 (9x more expensive than Qwen3.5 Plus!)
  • Context: ~200K tokens
  • Quality: ★★★★☆ (excellent)
  • Best For:
    • Multi-step reasoning
    • Complex planning
    • Algorithm design
    • Difficult debugging
  • When to Use: When reasoning/planning is required and budget models fail

GLM-5.1 — Maximum Quality

  • Model ID: glm-5.1
  • Cost: 880 requests per $12 (11.6x more expensive than Qwen3.5 Plus!)
  • Context: ~200K tokens
  • Quality: ★★★★★ (best available)
  • Speed: Moderate
  • Best For:
    • Critical architectural decisions
    • Complex multi-file refactoring
    • Production code review
    • When you need the absolute best quality
  • When to Use: Only when cheaper models can't handle the task

GLM-5.2 — Latest Premium Model

  • Model ID: glm-5.2
  • Cost: 880 requests per $12 (same as GLM-5.1)
  • Context: ~200K tokens
  • Quality: ★★★★★ (best available)
  • Speed: Moderate
  • Best For:
    • Latest GLM model with improvements over 5.1
    • Critical architectural decisions
    • Complex multi-file refactoring
    • Production code review
  • When to Use: Use instead of GLM-5.1 for the latest improvements

Kimi K3 — Latest Kimi Flagship

  • Model ID: kimi-k3
  • Provider: OpenCode Go (Moonshot AI upstream)
  • Endpoint: OpenAI-compatible (/v1/chat/completions)
  • Context: 1M tokens
  • Quality: ★★★★★
  • Max Output: 131K tokens
  • Modalities: Text, image, and video input
  • Cost: $3.00 / 1M input tokens · $15.00 / 1M output tokens
  • Released: July 2026
  • Best For:
    • Latest-generation code generation and agentic tool use
    • Long-context work (1M window) and very long outputs
    • Multimodal tasks (image/video input)
  • When to Use: When you want the newest Kimi generation; falls back to Kimi K2.7 Code, then Kimi K2.6

Kimi K2.7 Code — Code Specialist

  • Model ID: kimi-k2.7-code
  • Cost: 1,350 requests per $12
  • Context: ~256K tokens
  • Quality: ★★★★★ (excellent for code tasks)
  • Max Output: 32K tokens (highest available!)
  • Modalities: Text and image input
  • Speed: Fast
  • Best For:
    • Large code generation tasks
    • Complex refactoring requiring long outputs
    • Code review with detailed feedback
    • When you need the highest output token limit
  • When to Use: When you need both high quality AND very long outputs (up to 32K)

Qwen3.7 Plus — Upgraded General Coding

  • Model ID: qwen3.7-plus
  • Endpoint: Anthropic-compatible (/v1/messages)
  • Cost: 4,300 requests per $12 (better value than Qwen3.6!)
  • Context: ~1M tokens
  • Quality: ★★★★☆
  • Modalities: Text and image input
  • Speed: Fast
  • Best For:
    • General coding with better quality than Qwen3.6
    • Feature implementation
    • Bug fixes
  • When to Use: When you want better quality than Qwen3.6 at similar speed

Qwen3.7 Max — Maximum Quality Qwen

  • Model ID: qwen3.7-max
  • Endpoint: Anthropic-compatible (/v1/messages)
  • Cost: 950 requests per $12
  • Context: ~1M tokens
  • Quality: ★★★★☆
  • Modalities: Text and image input
  • Best For:
    • Complex coding tasks
    • When Qwen3.7 Plus isn't enough
  • When to Use: When you need Qwen's best quality

Usage Limits

OpenCode Go limits:

  • 5-hour limit: $12 of usage
  • Weekly limit: $30 of usage
  • Monthly limit: $60 of usage

Cost Comparison Example

Scenario: You want to make 5,000 requests this month.

ModelCostCan you do it?
Qwen3.5 Plus~$6✅ Yes, easily
MiniMax M2.5~$10✅ Yes
Qwen3.6 Plus~$18✅ Yes
Kimi K2.5~$32❌ Exceeds $30 weekly
GLM-5~$52❌ Exceeds limits
GLM-5.1~$68❌ Exceeds limits

Optimizing Your Usage

Strategy 1: Tiered Approach

1. Start with Qwen3.6 Plus (cheap, good quality)
2. If it fails, try Kimi K2.5 (better quality)
3. If still failing, use GLM-5 (reasoning)
4. Only for critical tasks: GLM-5.1 (premium)

Strategy 2: Task-Based Selection

Background ops (grep, ls, cat) → Qwen3.5 Plus
General coding → Qwen3.6 Plus or Kimi K2.5
Complex features → Kimi K2.5
Architecture/Planning → GLM-5
Critical review → GLM-5.1 (rarely)

Fallback Chains for Cost Efficiency

{
  "fallbacks": {
    "background": [
      { "model_id": "qwen3.6-plus" },
      { "model_id": "minimax-m2.5" }
    ],
    "long_context": [
      { "provider": "opencode-go", "model_id": "qwen3.7-plus" },
      { "provider": "opencode-go", "model_id": "qwen3.7-max" },
      { "provider": "opencode-zen", "model_id": "nemotron-3-ultra-free" },
      { "provider": "opencode-zen", "model_id": "mimo-v2.5-free" },
      { "provider": "opencode-zen", "model_id": "deepseek-v4-flash-free" }
    ],
    "default": [{ "model_id": "mimo-v2.5-pro" }, { "model_id": "qwen3.6-plus" }],
    "think": [{ "model_id": "kimi-k2.6" }],
    "complex": [{ "model_id": "glm-5" }],
    "fast": [{ "model_id": "qwen3.5-plus" }, { "model_id": "minimax-m2.5" }]
  }
}

Rule of thumb: If a task succeeds with a cheap model, it doesn't need an expensive one. Only fall back to expensive models when necessary.

Quick Reference

Task TypeRecommendedCost (req/$12)Fallback
Read file, ls, grepQwen3.5 Plus10,200Qwen3.6 Plus
General codingQwen3.7 Plus4,300Qwen3.6 Plus
Complex featuresKimi K2.61,850MiMo-V2.5-Pro
Long context (>100K)MiniMax M33,200Qwen3.7 Plus
Reasoning/planningGLM-51,150Kimi K2.6
Critical architectureGLM-5.2880GLM-5.1
Code specialistKimi K2.7 Code1,350Kimi K2.6
Bulk operationsQwen3.5 Plus10,200MiniMax M2.5

Cost-Saving Tips

  1. Use Qwen3.6 Plus as default — 3,300 req/$12 is plenty for most tasks
  2. Reserve GLM-5.1 for critical tasks only — 880 req/$12 drains budget fast
  3. Use Qwen3.5 Plus for simple operations — 10,200 req/$12 is unbeatable
  4. MiniMax M3 for long context — 3,200 req/$12 with a 1M window; MiniMax M2.5 stays the budget pick at 6,300 req/$12 as long as you fit inside its 200K window
  5. Use Zen free-tier models for non-critical tasks — Nemotron 3 Ultra Free, MiMo V2.5 Free, DeepSeek V4 Flash Free, Big Pickle, and others cost $0 while their promotions remain active
  6. Monitor your usage in the OpenCode console

See Also