Model comparison
Llama 3.1-8B vs MiniMax-M3
MiniMax-M3 is the stronger model overall, scoring 43.8 to 23.0 on the Noometry Index. Llama 3.1-8B costs 9.1× less per token, which makes it the better buy when MiniMax-M3's lead doesn't matter for your workload.
Last verified . 25 shared benchmarks.
Meta
MiniMaxSummary
- They share 25 benchmarks with published results for both. Llama 3.1-8B scores higher in 0 categories and MiniMax-M3 in 9 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in knowledge, where MiniMax-M3 leads 58.4 to 8.0.
- The biggest single-benchmark swing is OTIS Mock AIME 2024-2025: 1.7% for Llama 3.1-8B and 71.1% for MiniMax-M3.
- Llama 3.1-8B is cheaper at $0.05 / $0.08 per million input/output tokens, against $0.30 / $1.20 for MiniMax-M3.
- MiniMax-M3 accepts more context: 1M tokens versus 128K.
Side by side
| Llama 3.1-8B | MiniMax-M3 | |
|---|---|---|
| Provider | Meta | MiniMax |
| Noometry Index | 23.0 | 43.8 |
| Released | 2024-07-23 | 2026-06-01 |
| Weights | Open | Open |
| Context window | 128K | 1M |
| Max output | 4K | 512K |
| Input $ / M tokens | $0.05 | $0.30 |
| Output $ / M tokens | $0.08 | $1.20 |
| Results tracked | 43 | 41 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding MiniMax-M3 leads
Llama 3.1-8B: 20.2 (#340), MiniMax-M3: 41.8 (#118)
| Benchmark | Llama 3.1-8B | MiniMax-M3 |
|---|---|---|
| SciCode | 13.2% | 47.1% |
| LMArena Coding | 1195 | 1469 |
| FrontierCode | — | 14.7% |
| LMArena WebDev | — | 1482 |
| WeirdML | 1.7% | — |
| BigCodeBench Instruct | 32.8% | — |
| BigCodeBench Complete | 40.5% | — |
| ALE-Bench | — | 640.02 |
| HumanEval+ | 62.8% | — |
| MBPP+ | 55.6% | — |
Agentic & Tool Use Too close to call
Llama 3.1-8B: 22.5 (#131), MiniMax-M3: 22.6 (#130)
| Benchmark | Llama 3.1-8B | MiniMax-M3 |
|---|---|---|
| APEX-Agents | — | 37.7% |
| Berkeley Function Calling Leaderboard | 25.8% | — |
| OSWorld 2.0 | — | 4.6% |
| BALROG | 15.1% | — |
| GBAEval | — | 0.9% |
| Vending-Bench 2 | — | 2,158 |
Reasoning MiniMax-M3 leads
Llama 3.1-8B: 14.9 (#321), MiniMax-M3: 30.1 (#87)
| Benchmark | Llama 3.1-8B | MiniMax-M3 |
|---|---|---|
| CritPt | 0% | 3.7% |
| Chess Puzzles | 0% | 14% |
| LMArena Hard Prompts | 1175 | 1447 |
| DTBench | 50.9% | 78.9% |
| LMCA | 5.4% | 33.7% |
| Epoch Capabilities Index | 116.57 | 146.95 |
| SimpleBench | — | 45.8% |
| NYT Connections (extended) | — | 65.1% |
| Mystery Game Puzzles | — | 8% |
| Surface Evolver Bench | — | 55% |
| ForecastBench | — | 61.4 |
| PIQA | 81.2% | — |
Math MiniMax-M3 leads
Llama 3.1-8B: 10.2 (#317), MiniMax-M3: 40.0 (#95)
| Benchmark | Llama 3.1-8B | MiniMax-M3 |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 1.7% | 71.1% |
| LMArena Math | 1179 | 1429 |
| ProofBench | — | 18% |
| Omni-MATH | 13.7% | — |
| MATH Level 5 | 22.9% | — |
| GSM8K | 82.4% | — |
Knowledge MiniMax-M3 leads
Llama 3.1-8B: 8.0 (#307), MiniMax-M3: 58.4 (#35)
| Benchmark | Llama 3.1-8B | MiniMax-M3 |
|---|---|---|
| GPQA Diamond | 27% | 90.9% |
| LMArena Expert | 1144 | 1461 |
| MMLU-Pro | 40.6% | — |
| GPQA (HELM) | 24.7% | — |
| BoolQ | 82.8% | — |
| MMLU | 56.1% | — |
Multimodal Not comparable
Llama 3.1-8B: —, MiniMax-M3: 40.2 (#51)
| Benchmark | Llama 3.1-8B | MiniMax-M3 |
|---|---|---|
| LMArena Vision | — | 1253 |
| LMArena Document | — | 1435 |
Multilingual MiniMax-M3 leads
Llama 3.1-8B: 34.0 (#249), MiniMax-M3: 53.0 (#75)
| Benchmark | Llama 3.1-8B | MiniMax-M3 |
|---|---|---|
| LMArena Non-English | 1148 | 1420 |
| LMArena Chinese | 1151 | 1463 |
| LMArena French | 1177 | 1447 |
| LMArena German | 1144 | 1426 |
| LMArena Japanese | 1061 | 1381 |
| LMArena Korean | 1053 | 1372 |
| LMArena Russian | 1158 | 1428 |
| LMArena Spanish | 1169 | 1432 |
Instruction Following MiniMax-M3 leads
Llama 3.1-8B: 58.9 (#258), MiniMax-M3: 75.5 (#62)
| Benchmark | Llama 3.1-8B | MiniMax-M3 |
|---|---|---|
| LMArena Instruction Following | 1159 | 1433 |
| IFEval | 74.3% | — |
Long Context MiniMax-M3 leads
Llama 3.1-8B: 35.8 (#238), MiniMax-M3: 44.2 (#72)
| Benchmark | Llama 3.1-8B | MiniMax-M3 |
|---|---|---|
| LMArena Longer Query | 1182 | 1445 |
Writing & Preference MiniMax-M3 leads
Llama 3.1-8B: 29.7 (#290), MiniMax-M3: 62.1 (#83)
| Benchmark | Llama 3.1-8B | MiniMax-M3 |
|---|---|---|
| LMArena Text | 1187 | 1433 |
| LMArena Creative Writing | 1154 | 1404 |
| LMArena Multi-Turn | 1172 | 1442 |
| EQ-Bench Creative Writing | 713 | — |
| WildBench | 68.7% | — |
| EQ-Bench 4 | — | 1150 |
Frequently asked questions
Is Llama 3.1-8B better than MiniMax-M3?
MiniMax-M3 is the stronger model overall, scoring 43.8 to 23.0 on the Noometry Index. Llama 3.1-8B costs 9.1× less per token, which makes it the better buy when MiniMax-M3's lead doesn't matter for your workload.
Which is cheaper, Llama 3.1-8B or MiniMax-M3?
Llama 3.1-8B is cheaper. It lists at $0.05 per million input tokens and $0.08 per million output tokens; MiniMax-M3 lists at $0.30 and $1.20.
Is Llama 3.1-8B or MiniMax-M3 better for coding?
MiniMax-M3 scores higher on coding benchmarks: 41.8 versus 20.2 in the Noometry coding category.
Which has the bigger context window?
MiniMax-M3 does, with 1M tokens against 128K.
How many benchmarks do Llama 3.1-8B and MiniMax-M3 share?
25 benchmarks have published results for both models. Llama 3.1-8B has 43 scored results on Noometry and MiniMax-M3 has 41.