Chat model comparison
Benchmarks and pricing for models available through Meridian. Updated May 2026.
| Model | Context | MMLU-Pro P50 | $/M in | $/M out | Best for |
|---|---|---|---|---|---|
| GPT-5 | 256K | 92.3 | $2.50 | $10.00 | Complex reasoning, multi-step agents, code generation |
| Claude Opus 4.8 | 200K | 91.7 | $15.00 | $75.00 | Long-form analysis, nuanced instruction following, safety-critical |
| Gemini 3.1 Pro | 1M | 89.4 | $1.25 | $5.00 | Massive context windows, multimodal, cost-efficient throughput |
| Gemini Chrome | 32K | 84.1 | Free | Free | Browser-native tasks, quick drafts, zero-cost experimentation |
| Llama 4 Maverick | 128K | 87.6 | $0.20 | $0.60 | Self-hosted deployments, fine-tuning, budget-sensitive scale |
P50 scores from MMLU-Pro public leaderboard. Pricing reflects pay-as-you-go tiers. Gemini Chrome is available at no cost within usage limits.