Tool calling models, ranked
An agent lives or dies on two things: calling the right function with valid arguments, and returning JSON your code can parse without a retry. Everything here does both. In a multi-step loop, a cheaper model with a low error rate beats an expensive one with the same error rate, because every retry pays twice.
Most used in this category
Each model's share of this category's usage over the last 30 days, and its output price per 1M tokens in toman.
- 1
Gemini 3.8 Flash50% share1,298,323 toman - 2Gemini 2.5 Flash Lite29% share138,488 toman
- 3Gemini 3.1 Flash Lite5.6% share519,329 toman
- 4Claude Sonnet 54.6% share3,462,194 toman
- 5Gemini 2.5 Flash4.1% share865,548 toman
Best value in this category
Score out of 100: 50% cheapness, 30% capabilities, 20% context size — not a quality benchmark.
| # | Model | Context (tokens) | ≈ 1,000 words (toman) | Score |
|---|---|---|---|---|
| 1 | 1GLM Flash Latest | 1,048,576 | 54 | 86.6 |
| 2 | 2Qwen3.7 Flash | 1,000,000 | 72 | 83.3 |
| 3 | 3GPT-6 Luna Pro (batch) | 1,050,000 | 135 | 82.6 |
| 4 | 4GPT-6 Luna (batch) | 1,050,000 | 135 | 82.6 |
| 5 | 5Claude Haiku 5.5 (batch) | 1,000,000 | 135 | 82.5 |
| 6 | 6Ling 3.0 Tiny | 262,144 | 27 | 82.4 |
| 7 | 7GPT-5 Nano (batch) | 400,000 | 101 | 82.4 |
| 8 | 8Ling 3.0 Flash | 262,144 | 38 | 79.6 |
| 9 | 9GLM 5.3 Flash (batch) | 1,048,576 | 104 | 79.5 |
| 10 | 10Gemma 4 26B A4B | 262,144 | 67 | 79.4 |
| 11 | 11Gemini 2.5 Flash Lite | 1,048,576 | 225 | 77.1 |
| 12 | 12Ternary Bonsai 2 9B | 262,144 | 95 | 75.6 |
| 13 | 13Qwen3.5-Flash | 1,000,000 | 146 | 75.6 |
| 14 | 14GPT-6 Luna | 1,050,000 | 270 | 75.1 |
| 15 | 15GPT-6 Luna Pro | 1,050,000 | 270 | 75.1 |
Full list of models that call your tools reliably with prices →