Model comparison
Llama 3.1-8B vs Qwen3.5 27B
Qwen3.5 27B is the stronger model overall, scoring 41.9 to 23.0 on the Noometry Index. Llama 3.1-8B costs 14× less per token, which makes it the better buy when Qwen3.5 27B's lead doesn't matter for your workload.
Last verified . 20 shared benchmarks.
Meta
Alibaba (Qwen)Summary
- They share 20 benchmarks with published results for both. Llama 3.1-8B scores higher in 0 categories and Qwen3.5 27B in 8 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in knowledge, where Qwen3.5 27B leads 38.0 to 8.0.
- The biggest single-benchmark swing is WeirdML: 1.7% for Llama 3.1-8B and 39.5% for Qwen3.5 27B.
- Llama 3.1-8B is cheaper at $0.05 / $0.08 per million input/output tokens, against $0.30 / $2.40 for Qwen3.5 27B.
- Qwen3.5 27B accepts more context: 262K tokens versus 128K.
Side by side
| Llama 3.1-8B | Qwen3.5 27B | |
|---|---|---|
| Provider | Meta | Alibaba (Qwen) |
| Noometry Index | 23.0 | 41.9 |
| Released | 2024-07-23 | 2026-02-23 |
| Weights | Open | Open |
| Context window | 128K | 262K |
| Max output | 4K | 66K |
| Input $ / M tokens | $0.05 | $0.30 |
| Output $ / M tokens | $0.08 | $2.40 |
| Results tracked | 43 | 28 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Qwen3.5 27B leads
Llama 3.1-8B: 20.2 (#340), Qwen3.5 27B: 38.9 (#168)
| Benchmark | Llama 3.1-8B | Qwen3.5 27B |
|---|---|---|
| WeirdML | 1.7% | 39.5% |
| LMArena Coding | 1195 | 1427 |
| LMArena WebDev | — | 1358 |
| SciCode | 13.2% | — |
| BigCodeBench Instruct | 32.8% | — |
| BigCodeBench Complete | 40.5% | — |
| ALE-Bench | — | 349.45 |
| HumanEval+ | 62.8% | — |
| MBPP+ | 55.6% | — |
Agentic & Tool Use Not comparable
Llama 3.1-8B: 22.5 (#131), Qwen3.5 27B: —
| Benchmark | Llama 3.1-8B | Qwen3.5 27B |
|---|---|---|
| Berkeley Function Calling Leaderboard | 25.8% | — |
| BALROG | 15.1% | — |
| Vending-Bench 2 | — | 201.98 |
Reasoning Qwen3.5 27B leads
Llama 3.1-8B: 14.9 (#321), Qwen3.5 27B: 27.5 (#117)
| Benchmark | Llama 3.1-8B | Qwen3.5 27B |
|---|---|---|
| LMArena Hard Prompts | 1175 | 1414 |
| DTBench | 50.9% | 82.4% |
| LMCA | 5.4% | 34% |
| NYT Connections (extended) | — | 47.9% |
| CritPt | 0% | — |
| Chess Puzzles | 0% | — |
| Thematic Generalization | — | 45.5% |
| Epoch Capabilities Index | 116.57 | — |
| PIQA | 81.2% | — |
Math Qwen3.5 27B leads
Llama 3.1-8B: 10.2 (#317), Qwen3.5 27B: 38.8 (#127)
| Benchmark | Llama 3.1-8B | Qwen3.5 27B |
|---|---|---|
| LMArena Math | 1179 | 1429 |
| MathArena Final-Answer Competitions | — | 56.7% |
| OTIS Mock AIME 2024-2025 | 1.7% | — |
| Omni-MATH | 13.7% | — |
| MATH Level 5 | 22.9% | — |
| GSM8K | 82.4% | — |
Knowledge Qwen3.5 27B leads
Llama 3.1-8B: 8.0 (#307), Qwen3.5 27B: 38.0 (#150)
| Benchmark | Llama 3.1-8B | Qwen3.5 27B |
|---|---|---|
| LMArena Expert | 1144 | 1428 |
| GPQA Diamond | 27% | — |
| MMLU-Pro | 40.6% | — |
| Vectara Hallucination Rate | — | 12.1% |
| GPQA (HELM) | 24.7% | — |
| BoolQ | 82.8% | — |
| MMLU | 56.1% | — |
Multimodal Not comparable
Llama 3.1-8B: —, Qwen3.5 27B: 39.4 (#59)
| Benchmark | Llama 3.1-8B | Qwen3.5 27B |
|---|---|---|
| LMArena Vision | — | 1241 |
Multilingual Qwen3.5 27B leads
Llama 3.1-8B: 34.0 (#249), Qwen3.5 27B: 50.8 (#115)
| Benchmark | Llama 3.1-8B | Qwen3.5 27B |
|---|---|---|
| LMArena Non-English | 1148 | 1390 |
| LMArena Chinese | 1151 | 1478 |
| LMArena French | 1177 | 1410 |
| LMArena German | 1144 | 1393 |
| LMArena Japanese | 1061 | 1345 |
| LMArena Korean | 1053 | 1358 |
| LMArena Russian | 1158 | 1390 |
| LMArena Spanish | 1169 | 1407 |
Instruction Following Qwen3.5 27B leads
Llama 3.1-8B: 58.9 (#258), Qwen3.5 27B: 73.5 (#119)
| Benchmark | Llama 3.1-8B | Qwen3.5 27B |
|---|---|---|
| LMArena Instruction Following | 1159 | 1393 |
| IFEval | 74.3% | — |
Long Context Qwen3.5 27B leads
Llama 3.1-8B: 35.8 (#238), Qwen3.5 27B: 43.1 (#106)
| Benchmark | Llama 3.1-8B | Qwen3.5 27B |
|---|---|---|
| LMArena Longer Query | 1182 | 1413 |
Writing & Preference Qwen3.5 27B leads
Llama 3.1-8B: 29.7 (#290), Qwen3.5 27B: 59.3 (#111)
| Benchmark | Llama 3.1-8B | Qwen3.5 27B |
|---|---|---|
| LMArena Text | 1187 | 1409 |
| LMArena Creative Writing | 1154 | 1362 |
| LMArena Multi-Turn | 1172 | 1410 |
| EQ-Bench Creative Writing | 713 | — |
| WildBench | 68.7% | — |
Frequently asked questions
Is Llama 3.1-8B better than Qwen3.5 27B?
Qwen3.5 27B is the stronger model overall, scoring 41.9 to 23.0 on the Noometry Index. Llama 3.1-8B costs 14× less per token, which makes it the better buy when Qwen3.5 27B's lead doesn't matter for your workload.
Which is cheaper, Llama 3.1-8B or Qwen3.5 27B?
Llama 3.1-8B is cheaper. It lists at $0.05 per million input tokens and $0.08 per million output tokens; Qwen3.5 27B lists at $0.30 and $2.40.
Is Llama 3.1-8B or Qwen3.5 27B better for coding?
Qwen3.5 27B scores higher on coding benchmarks: 38.9 versus 20.2 in the Noometry coding category.
Which has the bigger context window?
Qwen3.5 27B does, with 262K tokens against 128K.
How many benchmarks do Llama 3.1-8B and Qwen3.5 27B share?
20 benchmarks have published results for both models. Llama 3.1-8B has 43 scored results on Noometry and Qwen3.5 27B has 28.