All Coding HarnessesVerified: 2026-11-19
Open-Source Cross-IDE & TUI Agent227.9kOpenCode v1.19.6 / Continue v2.0.0Apache-2.0 / MIT

OpenCode / Continue: Open BYOK Cost Optimization & Token Audit

Vendor-neutral open-source agent ecosystem (OpenCode TUI + Continue VS Code/JetBrains) via declarative config.yaml/json.

curl -fsSL https://opencode.ai/install | bash # or install Continue in VS Code/JetBrains
Install Friction
★★☆☆☆
2/5 · CLI / IDE
Runtime Dependency
VS Code / JetBrains IDE (Continue) or macOS/Linux Terminal (OpenCode)Git
BYOK & BaseURL
Full Support5 Env Vars
Typical Multiplier
1.5x – 2.3xTypical 1.9x vs Raw API

Context & Token Overhead Mechanics

In a 15-turn refactor with 5 tool calls, explicit @-file scoping and a slim 1.85k system prompt keep overhead at 1.5x–2.3x.

15-Turn Benchmark ≈ 55,500 Input Tokens
Base Task Payload:30,000 (54%)
System Prompt Overhead:3,700 (7%)
Context / Repo-Map:9,810 (18%)
Multi-Turn Tool Loop:11,990 (21%)
System Prompt Overhead
~1,850 tokens

Baseline instructions & tool schema weight per turn

Context Compression

Explicit @-context providers + OpenCode /compact session pruning + minimal 1.8k system prompt

Prompt Caching Hit Rate72%–86%

Supports prompt caching natively on Anthropic, DeepSeek, and OpenRouter endpoints when `cache: true` / `promptCaching` is enabled.

Model Compatibility Matrix & 15-Turn Cost

Estimated on 30k raw input + 4.5k output scaled by 1.9x multiplier at 70% cache hit rate

ModelTool CallingPrompt CachingVisionStreamingEst. 15-Turn Cost
DeepSeek-V3 (deepseek-chat)Primary Open-SourceNativeFullN/AStable$0.014
DeepSeek: DeepSeek V4.1 FlashUltra-Low CostNativeFullSupportedStable$0.002
Llama 3.3 70B Versatile420 t/s Autocomplete & Fast ChatNativeFullN/AStable$0.026
Google: Gemini 4 NeonFrontier 1M AgentNativeFullSupportedStable$0.146

Production BYOK & DeepSeek / OpenRouter Config

Zero vendor lock-in: configure OpenRouter, DeepSeek, Groq LPU, LiteLLM, vLLM, or Ollama declaratively in `opencode.json` or `~/.continue/config.yaml`.

{
  "$schema": "https://opencode.ai/config.json",
  "model": "deepseek/deepseek-chat",
  "small_model": "openrouter/deepseek/deepseek-v4.1-flash",
  "provider": {
    "deepseek": { "options": { "apiKey": "{env:DEEPSEEK_API_KEY}" } },
    "openrouter": { "options": { "apiKey": "{env:OPENROUTER_API_KEY}" } }
  }
}
Supported Env Vars:OPENROUTER_API_KEYDEEPSEEK_API_KEYOPENAI_API_KEYANTHROPIC_API_KEYGROQ_API_KEY

OpenCode / Continue Cost Pitfalls & Tuning FAQ

Tool Calling in Continue/OpenCode requires native `functionCalling` support. Verify the model supports `capabilities: [tool_use]` (e.g., DeepSeek-V3, Gemini 4 Neon, GPT-6.2 Sol) instead of raw reasoning-only checkpoints.

Compare Other Coding Harnesses

All 7 Benchmarks