Compare providers
Pick up to three. Scores are curated estimates on a 0–10 scale. Lock-in: higher = worse.
| Dimension | OpenAI | Anthropic | Google AI (Gemini) |
|---|---|---|---|
| Overall | 8.2 | 8.4 | 8.4 |
| Capability fit | 9.6 | 9.7 | 9.2 |
| Price / performance | 6.2 | 6.8 | 8.8 |
| Latency / reliability | 8.8 | 8.7 | 8.5 |
| Developer & agent DX | 9.5 | 9.2 | 8.2 |
| Privacy / data retention | 7.8 | 9.0 | 8.0 |
| Lock-in risk (higher = worse) | 7.2 | 6.5 | 6.0 |
| Maturity / track record | 9.8 | 9.4 | 9.6 |
| Pricing | GPT-5.6 tiers (Sol/Terra/Luna) roughly $5/$30 to $1/$6 per 1M tokens; prompt caching with explicit breakpoints; batch discounts. Media (images/audio) billed separately. | Sonnet 5 ~$2/$10 per 1M; Opus ~$5/$25; Haiku 4.5 ~$1/$5; Fable higher. Prompt caching + batch. Managed Agents session-hour fees. US-only inference ~1.1x. | Gemini 3.x Pro often ~$2/$12 under 200K; Flash tiers much cheaper. 1M context. Vertex for enterprise. Batch half-price common. |