Qwen2.5 7B Instruct vs Qwen3.5-35B-A3B
Qwen2.5 7B Instruct costs 6.5x less per output token than Qwen3.5-35B-A3B ($0.20 vs $1.30 per million). Updated October 22, 2026; 5 benchmarks have a result for both.
Qwen2.5 7B Instruct or Qwen3.5-35B-A3B: which to pick
Neither model has a coding score yet, so the two cannot be ranked on coding.
Pick Qwen2.5 7B Instruct for output tokens 6.5x cheaper ($0.20 vs $1.30 per million) and work measured by Epoch Capabilities Index or GPQA Diamond (Epoch AI run), where it leads.
- Shared results
- 5
- Qwen2.5 7B Instruct only
- 2
- Qwen3.5-35B-A3B only
- 40
- Like-for-like categories
- 4 of 9
Only benchmarks that tested both models count toward a pick; a result only one model has is shown but not compared.
Which one for your work
Each card is decided by one number both models carry. Where only one has it, the card says so and makes no call.
Coding
Vibe Coding Index, six public coding benchmarks
No call
Neither has a coding score yet.
One side not measured
Intelligence
AA Intelligence Index
No call
Only Qwen3.5-35B-A3B has a AA Intelligence Index result, 19.3.
Directional only
Long documents
Listed context window
Qwen3.5-35B-A3B
Qwen3.5-35B-A3B takes 262K tokens against 33K.
Listed by OpenRouter
Everyday cost
1K input + 500 output tokens
Qwen2.5 7B Instruct
Qwen2.5 7B Instruct $0.0002, Qwen3.5-35B-A3B $0.0008 per turn.
List prices
At a glance
- Coding index
- Qwen2.5 7B Instruct not benchmarked, Qwen3.5-35B-A3B not benchmarked.
- Widest category
- Math, 67.5 points apart on average over 1 shared benchmark.
- Widest single gap
- OTIS Mock AIME: Qwen2.5 7B Instruct 2.5% against Qwen3.5-35B-A3B 70.0%, 67.5 points apart on a 0 to 100 scale.
- Output price
- Qwen2.5 7B Instruct $0.20 against Qwen3.5-35B-A3B $1.30 per million output tokens, 6.5x apart.
- Context window
- Qwen2.5 7B Instruct takes 33K tokens, Qwen3.5-35B-A3B 262K.
Benchmark by benchmark
First the score: boards each model leads, per category, where both have a result. Then every board, a category to a fold; bars are each result on a 0 to 100 scale, so compare the two in a row, not one row with another.
| Category | Qwen2.5 7B Instruct | Bars | Qwen3.5-35B-A3B | Ties |
|---|---|---|---|---|
| Intelligence | 0 | 1 | ||
| Reasoning | 0 | 1 | ||
| Math | 0 | 1 | ||
| General | 0 | 2 | ||
| All shared boards | 0 | 5 |
CodingDirectional onlySix of these boards feed the Vibe Coding Index7 boards, none shared
| Benchmark | Qwen2.5 7B Instruct | Qwen3.5-35B-A3B | Leader |
|---|---|---|---|
| LMArena WebDev | Not measured | 1254 | Only Qwen3.5-35B-A3B measured |
| Terminal-Bench Hard (Artificial Analysis run) | Not measured | 26.5% | Only Qwen3.5-35B-A3B measured |
| ALE-Bench | Not measured | 221.8 | Only Qwen3.5-35B-A3B measured |
| LMArena Text: coding | Not measured | 1410 | Only Qwen3.5-35B-A3B measured |
| LMArena Text: software and IT | Not measured | 1415 | Only Qwen3.5-35B-A3B measured |
| LMArena WebDev: React | Not measured | 1238 | Only Qwen3.5-35B-A3B measured |
| LMArena WebDev: HTML | Not measured | 1301 | Only Qwen3.5-35B-A3B measured |
| Neither model has a result on the other 32 boards here | |||
IntelligenceLike-for-likePublic boards, shown as published, kept out of the index01
| Benchmark | Qwen2.5 7B Instruct | Qwen3.5-35B-A3B | Leader |
|---|---|---|---|
| AA Intelligence Index | Not measured | 19.3 | Only Qwen3.5-35B-A3B measured |
| Epoch Capabilities Index | 118.5 | 142.5 | Qwen3.5-35B-A3B |
Agentic workDirectional onlyPublic boards, shown as published, kept out of the index3 boards, none shared
| Benchmark | Qwen2.5 7B Instruct | Qwen3.5-35B-A3B | Leader |
|---|---|---|---|
| τ²-Bench Telecom | Not measured | 89.2% | Only Qwen3.5-35B-A3B measured |
| BALROG | 7.8% | Not measured | Only Qwen2.5 7B Instruct measured |
| Vending-Bench 2 | Not measured | $463 | Only Qwen3.5-35B-A3B measured |
| Neither model has a result on the other 38 boards here | |||
ReasoningLike-for-likePublic boards, shown as published, kept out of the index01
| Benchmark | Qwen2.5 7B Instruct | Qwen3.5-35B-A3B | Leader |
|---|---|---|---|
| GPQA Diamond (Artificial Analysis run) | Not measured | 84.5% | Only Qwen3.5-35B-A3B measured |
| Humanity's Last Exam (Artificial Analysis run) | Not measured | 21.0% | Only Qwen3.5-35B-A3B measured |
| GPQA Diamond (Epoch AI run) | 35.5% | 83.5% | Qwen3.5-35B-A3B |
| LMArena Text: hard prompts | Not measured | 1400 | Only Qwen3.5-35B-A3B measured |
| LMArena Text: expert | Not measured | 1408 | Only Qwen3.5-35B-A3B measured |
| LMArena Text: hard prompts in English | Not measured | 1411 | Only Qwen3.5-35B-A3B measured |
| Neither model has a result on the other 16 boards here | |||
MathLike-for-likePublic boards, shown as published, kept out of the index01
| Benchmark | Qwen2.5 7B Instruct | Qwen3.5-35B-A3B | Leader |
|---|---|---|---|
| OTIS Mock AIME | 2.5% | 70.0% | Qwen3.5-35B-A3B |
| LMArena Text: math | Not measured | 1404 | Only Qwen3.5-35B-A3B measured |
| LMArena Text: mathematical work | Not measured | 1410 | Only Qwen3.5-35B-A3B measured |
| Neither model has a result on the other 10 boards here | |||
KnowledgeDirectional onlyPublic boards, shown as published, kept out of the index5 boards, none shared
| Benchmark | Qwen2.5 7B Instruct | Qwen3.5-35B-A3B | Leader |
|---|---|---|---|
| MMLU | 72.9% | Not measured | Only Qwen2.5 7B Instruct measured |
| LMArena Text: science | Not measured | 1413 | Only Qwen3.5-35B-A3B measured |
| LMArena Text: medicine and healthcare | Not measured | 1421 | Only Qwen3.5-35B-A3B measured |
| LMArena Text: legal and government | Not measured | 1399 | Only Qwen3.5-35B-A3B measured |
| LMArena Text: business and finance | Not measured | 1394 | Only Qwen3.5-35B-A3B measured |
| Neither model has a result on the other 13 boards here | |||
Web designNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
| Benchmark | Qwen2.5 7B Instruct | Qwen3.5-35B-A3B | Leader |
|---|---|---|---|
| Neither model has a result on 9 boards here | |||
GeneralLike-for-likePublic boards, shown as published, kept out of the index02
| Benchmark | Qwen2.5 7B Instruct | Qwen3.5-35B-A3B | Leader |
|---|---|---|---|
| IFBench | Not measured | 72.5% | Only Qwen3.5-35B-A3B measured |
| Long context reasoning (AA-LCR) | Not measured | 72.0% | Only Qwen3.5-35B-A3B measured |
| DTBench | 47.7% | 80.0% | Qwen3.5-35B-A3B |
| LMCA | 6.4% | 29.5% | Qwen3.5-35B-A3B |
| LMArena Text | Not measured | 1395 | Only Qwen3.5-35B-A3B measured |
| LMArena Text: instruction following | Not measured | 1379 | Only Qwen3.5-35B-A3B measured |
| LMArena Text: creative writing | Not measured | 1347 | Only Qwen3.5-35B-A3B measured |
| LMArena Text: English | Not measured | 1407 | Only Qwen3.5-35B-A3B measured |
| LMArena Text: non-English | Not measured | 1378 | Only Qwen3.5-35B-A3B measured |
| LMArena Text: Chinese | Not measured | 1457 | Only Qwen3.5-35B-A3B measured |
| LMArena Text: German | Not measured | 1367 | Only Qwen3.5-35B-A3B measured |
| LMArena Text: Japanese | Not measured | 1326 | Only Qwen3.5-35B-A3B measured |
| LMArena Text: Korean | Not measured | 1355 | Only Qwen3.5-35B-A3B measured |
| LMArena Text: French | Not measured | 1412 | Only Qwen3.5-35B-A3B measured |
| LMArena Text: Spanish | Not measured | 1390 | Only Qwen3.5-35B-A3B measured |
| LMArena Text: Russian | Not measured | 1376 | Only Qwen3.5-35B-A3B measured |
| LMArena Text: Polish | Not measured | 1385 | Only Qwen3.5-35B-A3B measured |
| LMArena Text: multi-turn | Not measured | 1390 | Only Qwen3.5-35B-A3B measured |
| LMArena Text: longer queries | Not measured | 1389 | Only Qwen3.5-35B-A3B measured |
| LMArena Text: writing and language | Not measured | 1365 | Only Qwen3.5-35B-A3B measured |
| LMArena Text: entertainment and media | Not measured | 1347 | Only Qwen3.5-35B-A3B measured |
| Neither model has a result on the other 25 boards here | |||
Image generationNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
| Benchmark | Qwen2.5 7B Instruct | Qwen3.5-35B-A3B | Leader |
|---|---|---|---|
| Neither model has a result on 5 boards here | |||
What each workload costs
Three fixed token mixes at each model's listed price, so per-token rates become a bill.
Chat turn
1K input + 500 output tokens
Qwen2.5 7B Instruct costs less
Repository review
50K input + 3K output tokens
Qwen2.5 7B Instruct costs less
Agent loop
200K input + 10K output tokens
Qwen2.5 7B Instruct costs less
Listed OpenRouter rates, without cache or batch discounts. A workload that does not fit a model's context window would need splitting.
Specification differences
The leading value in a row is marked; a row with a missing value names no leader. Our two rows are marked in red.
| Specification | Qwen2.5 7B Instruct | Qwen3.5-35B-A3B |
|---|---|---|
| Maker | Qwen | Qwen |
| Weights | Open weights | Open weights |
| Released | October 16, 2024 | February 25, 2026 |
| Context window | 33K | 262K |
| Input $/Mtok | $0.10 | $0.1625 |
| Output $/Mtok | $0.20 | $1.30 |
| Coding index | Not benchmarked | Not benchmarked |
| Independent measurements behind it | 0 | 1 |
| Cost per solved task | Not computable | Not computable |
| Source | OpenRouter listing | OpenRouter listing |
Questions
Is Qwen2.5 7B Instruct better than Qwen3.5-35B-A3B for coding?
Neither model has a coding score yet, so the two cannot be ranked on coding. Pick Qwen2.5 7B Instruct for output tokens 6.5x cheaper ($0.20 vs $1.30 per million) and work measured by Epoch Capabilities Index or GPQA Diamond (Epoch AI run), where it leads.
Which is cheaper, Qwen2.5 7B Instruct or Qwen3.5-35B-A3B?
Qwen2.5 7B Instruct costs $0.10 per million input tokens and $0.20 per million output tokens; Qwen3.5-35B-A3B costs $0.1625 and $1.30. Cost per solved task is known for neither of them.
Which has the bigger context window, Qwen2.5 7B Instruct or Qwen3.5-35B-A3B?
Qwen3.5-35B-A3B, with 262K tokens against 33K.
How many benchmarks is the Qwen2.5 7B Instruct vs Qwen3.5-35B-A3B comparison based on?
5 benchmarks have a result for both models as of October 22, 2026. Qwen2.5 7B Instruct's coding index rests on 0 independent benchmarks and Qwen3.5-35B-A3B's on 1. Where only one model has a result, the page says so instead of naming a winner.