Agents resend conversation state on every step and retry failed calls. Reliability matters as much as price — a cheap model with unreliable tool calls can cost more once retries are counted.
| # | Model | Provider | Input / 1M | Output / 1M | Context | Score | Profile |
|---|---|---|---|---|---|---|---|
| 1 | Claude Haiku 5.5 | Anthropic | $0.100 | $0.500 | 1.0M | 0.73 | View |
| 2 | Claude Sonnet 5.5 | Anthropic | $2.000 | $10.000 | 1.0M | 0.68 | View |
| 3 | Gemini 3.5 Flash-Lite | $0.300 | $2.500 | 1.0M | 0.68 | View | |
| 4 | DeepSeek Flash | DeepSeek | $0.150 | $0.600 | 1.0M | 0.66 | View |
| 5 | Gemini 3.1 Pro (Preview) | $2.000 | $12.000 | 1.0M | 0.64 | View | |
| 6 | Grok 4.3 | xAI | $1.250 | $2.500 | 1.0M | 0.63 | View |
| 7 | GPT-6.1 Sol | OpenAI | $2.000 | $10.000 | 1.1M | 0.63 | View |
| 8 | Claude Opus 5.5 | Anthropic | $4.000 | $20.000 | 1.0M | 0.62 | View |
| 9 | GPT-6 Luna | OpenAI | $0.100 | $0.500 | 1.1M | 0.61 | View |
| 10 | Gemini 3.8 Flash | $0.750 | $3.750 | 1.0M | 0.60 | View | |
| 11 | GPT-5.4 Mini | OpenAI | $0.750 | $4.500 | 400k | 0.60 | View |
| 12 | DeepSeek V4 Pro | DeepSeek | $0.660 | $1.980 | 1.0M | 0.56 | View |
| 13 | Gemini 4 Argon | $4.000 | $20.000 | 1.0M | 0.55 | View | |
| 14 | GPT-5.5 | OpenAI | $5.000 | $30.000 | 922k | 0.52 | View |
| 15 | Grok 4.7 | xAI | $2.000 | $6.000 | 500k | 0.45 | View |
Composite score = 0.25×cost + 0.2×context + 0.35×intelligence + 0.2×speed, where each factor is normalized 0–1 across the active catalog. Cost is inverted (lower price = higher score). Weights reflect the priorities of this use case. Scores are a starting point — always test on your own workload.
Models with strong tool-use reliability and structured output. Test on your own tool schemas before committing.
Each retry resends the full conversation state, multiplying input tokens. Reliability often beats raw price.