← All comparisons
Llama 3.3 Instruct 70B vs Claude 4.5 Haiku (Non-reasoning)
Meta vs Anthropic — side-by-side benchmark comparison
| Llama 3.3 Instruct 70B | Claude 4.5 Haiku (Non-reasoning) | |
|---|---|---|
| Intelligence Index | 14.5 | 31.0 |
| Coding Index | 10.7 | 29.6 |
| Math Index | 7.7 | 39.0 |
| Output speed (tok/s) | 88.1 | 105.8 |
| Blended price ($/1M) | $0.62 | $2.19 |
| Time to first token (s) | 0.59s | 0.65s |
| aime | 30.0% | — |
| aime 25 | 7.7% | 39.0% |
| artificial analysis coding index | 10.70 | 29.60 |
| artificial analysis intelligence index | 14.50 | 31.00 |
| artificial analysis math index | 7.70 | 39.00 |
| gpqa | 49.8% | 64.6% |
| hle | 4.0% | 4.3% |
| ifbench | 47.1% | 42.0% |
| lcr | 15.0% | 43.7% |
| livecodebench | 28.8% | 51.1% |
| math 500 | 77.3% | — |
| mmlu pro | 71.3% | 80.0% |
| scicode | 26.0% | 34.4% |
| tau2 | 26.6% | 32.5% |
| terminalbench hard | 3.0% | 27.3% |
Benchmark data from Artificial Analysis.