Model comparison
Qwen2.5 7B Instruct vs Qwen3.8 27B
Qwen3.8 27B is the stronger model overall, scoring 46.0 to 29.0 on the Noometry Index. Qwen2.5 7B Instruct costs 2.0× less per token, which makes it the better buy when Qwen3.8 27B's lead doesn't matter for your workload.
Last verified . 3 shared benchmarks.
Alibaba (Qwen)Summary
- They share 3 benchmarks with published results for both. Qwen2.5 7B Instruct scores higher in 0 categories and Qwen3.8 27B in 7 categories; 7 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where Qwen3.8 27B leads 41.0 to 14.8.
- The biggest single-benchmark swing is DTBench: 47.7% for Qwen2.5 7B Instruct and 88% for Qwen3.8 27B.
- Qwen2.5 7B Instruct is cheaper at $0.17 / $0.70 per million input/output tokens, against $0.04 / $2.30 for Qwen3.8 27B.
- Qwen3.8 27B accepts more context: 262K tokens versus 131K.
Side by side
| Qwen2.5 7B Instruct | Qwen3.8 27B | |
|---|---|---|
| Provider | Alibaba (Qwen) | Alibaba (Qwen) |
| Noometry Index | 29.0 | 46.0 |
| Released | 2024-09 | 2026-08-14 |
| Weights | Open | Open |
| Context window | 131K | 262K |
| Max output | 8K | 33K |
| Input $ / M tokens | $0.17 | $0.04 |
| Output $ / M tokens | $0.70 | $2.30 |
| Results tracked | 15 | 31 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Qwen3.8 27B leads
Qwen2.5 7B Instruct: 36.5 (#208), Qwen3.8 27B: 50.5 (#44)
| Benchmark | Qwen2.5 7B Instruct | Qwen3.8 27B |
|---|---|---|
| LMArena WebDev | — | 1593 |
| SciCode | — | 46.6% |
| BigCodeBench Instruct | 37.6% | — |
| LMArena Coding | — | 1482 |
| BigCodeBench Complete | 46.1% | — |
Agentic & Tool Use Qwen3.8 27B leads
Qwen2.5 7B Instruct: 23.8 (#124), Qwen3.8 27B: 32.9 (#57)
| Benchmark | Qwen2.5 7B Instruct | Qwen3.8 27B |
|---|---|---|
| APEX-Agents | — | 47.5% |
| BALROG | 7.8% | — |
Reasoning Qwen3.8 27B leads
Qwen2.5 7B Instruct: 14.8 (#322), Qwen3.8 27B: 41.0 (#54)
| Benchmark | Qwen2.5 7B Instruct | Qwen3.8 27B |
|---|---|---|
| DTBench | 47.7% | 88% |
| LMCA | 6.4% | 41.4% |
| Epoch Capabilities Index | 118.51 | 149.38 |
| ARC-AGI-2 | — | 42.4% |
| NYT Connections (extended) | — | 54.5% |
| ARC-AGI-1 | — | 87.5% |
| CritPt | — | 5.4% |
| Chess Puzzles | 0% | — |
| LMArena Hard Prompts | — | 1460 |
| Surface Evolver Bench | — | 45% |
Math Qwen3.8 27B leads
Qwen2.5 7B Instruct: 12.6 (#306), Qwen3.8 27B: 37.1 (#161)
| Benchmark | Qwen2.5 7B Instruct | Qwen3.8 27B |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 2.5% | — |
| ProofBench | — | 16% |
| Omni-MATH | 29.4% | — |
| LMArena Math | — | 1456 |
Knowledge Qwen3.8 27B leads
Qwen2.5 7B Instruct: 17.0 (#286), Qwen3.8 27B: 41.6 (#109)
| Benchmark | Qwen2.5 7B Instruct | Qwen3.8 27B |
|---|---|---|
| GPQA Diamond | 35.5% | — |
| MMLU-Pro | 53.9% | — |
| GPQA (HELM) | 34.1% | — |
| LMArena Expert | — | 1482 |
| MMLU | 72.9% | — |
Multimodal Not comparable
Qwen2.5 7B Instruct: —, Qwen3.8 27B: 41.3 (#37)
| Benchmark | Qwen2.5 7B Instruct | Qwen3.8 27B |
|---|---|---|
| LMArena Vision | — | 1271 |
Multilingual Not comparable
Qwen2.5 7B Instruct: —, Qwen3.8 27B: 53.7 (#60)
| Benchmark | Qwen2.5 7B Instruct | Qwen3.8 27B |
|---|---|---|
| LMArena Non-English | — | 1430 |
| LMArena Chinese | — | 1504 |
| LMArena French | — | 1465 |
| LMArena German | — | 1438 |
| LMArena Japanese | — | 1384 |
| LMArena Korean | — | 1393 |
| LMArena Russian | — | 1415 |
| LMArena Spanish | — | 1448 |
Instruction Following Qwen3.8 27B leads
Qwen2.5 7B Instruct: 63.2 (#231), Qwen3.8 27B: 75.8 (#53)
| Benchmark | Qwen2.5 7B Instruct | Qwen3.8 27B |
|---|---|---|
| IFEval | 74.1% | — |
| LMArena Instruction Following | — | 1439 |
Long Context Not comparable
Qwen2.5 7B Instruct: —, Qwen3.8 27B: 44.3 (#70)
| Benchmark | Qwen2.5 7B Instruct | Qwen3.8 27B |
|---|---|---|
| LMArena Longer Query | — | 1450 |
Writing & Preference Qwen3.8 27B leads
Qwen2.5 7B Instruct: 48.8 (#195), Qwen3.8 27B: 65.8 (#43)
| Benchmark | Qwen2.5 7B Instruct | Qwen3.8 27B |
|---|---|---|
| LMArena Text | — | 1441 |
| LMArena Creative Writing | — | 1384 |
| EQ-Bench Creative Writing | — | 1671 |
| WildBench | 73.1% | — |
| LMArena Multi-Turn | — | 1441 |
Frequently asked questions
Is Qwen2.5 7B Instruct better than Qwen3.8 27B?
Qwen3.8 27B is the stronger model overall, scoring 46.0 to 29.0 on the Noometry Index. Qwen2.5 7B Instruct costs 2.0× less per token, which makes it the better buy when Qwen3.8 27B's lead doesn't matter for your workload.
Which is cheaper, Qwen2.5 7B Instruct or Qwen3.8 27B?
Qwen2.5 7B Instruct is cheaper. It lists at $0.17 per million input tokens and $0.70 per million output tokens; Qwen3.8 27B lists at $0.04 and $2.30.
Is Qwen2.5 7B Instruct or Qwen3.8 27B better for coding?
Qwen3.8 27B scores higher on coding benchmarks: 50.5 versus 36.5 in the Noometry coding category.
Which has the bigger context window?
Qwen3.8 27B does, with 262K tokens against 131K.
How many benchmarks do Qwen2.5 7B Instruct and Qwen3.8 27B share?
3 benchmarks have published results for both models. Qwen2.5 7B Instruct has 15 scored results on Noometry and Qwen3.8 27B has 31.