LLM model APIs — compare pricing, reliability & setup
Generated 2026-11-03 03:53 UTC · JSON
For agentic coding, the current pick is Gemini 4 Argon (score 78.87). See cost by profile below.
Ranking — Agentic coding
| Vendor | Score | Coding (50M in / 10M out) | Bulk extract (100M in / 5M out) | Free | Agent-friction | Verified |
|---|---|---|---|---|---|---|
| Gemini 4 Argon | 78.87 | $200 | $250 | yes | 90 | 2026-11-02 |
| Claude Opus 5.5 | 76.89 | $400 | $500 | yes | 90 | 2026-11-02 |
| Gemini 3.8 Flash | 76.56 | $75 | $93.75 | yes | 90 | 2026-11-02 |
| Claude Haiku 5.5 | 76.5 | $10 | $12.50 | yes | 90 | 2026-11-02 |
| GPT-6 Luna | 75.8 | $10 | $12.50 | yes | 90 | 2026-11-02 |
| GPT-6 Astra | 67.1 | $1000 | $1250 | yes | 90 | 2026-11-02 |
Agent-friction — can a service be set up without a human?
Each check and its weight is published; the column total is the 0–100 agent-friction score.
| Check | Gemini 4 Argon | Claude Opus 5.5 | Gemini 3.8 Flash | Claude Haiku 5.5 | GPT-6 Luna | GPT-6 Astra |
|---|---|---|---|---|---|---|
| Test credentials without a credit card (20) | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ |
| Sandbox or test mode (15) | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ |
| Setup with no human-only steps (15) | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ |
| Official CLI (10) | × | × | × | × | × | × |
| Official MCP server (10) | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ |
| Markdown docs, current with the SDK (10) | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ |
| Typed SDK for main languages (10) | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ |
| Error messages that name the fix (10) | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ |
| Score | 90 | 90 | 90 | 90 | 90 | 90 |
How this category is scored
Per use case, weights published in the data: fit 35% · cost 15% · integration 15% · agent_friction 15% · health 10% · lock_in 10%. Affiliate status is never an input.