Qwen2.5 72B Instruct vs Qwen3 Coder Next
Updated October 21, 2026. 7 benchmarks have a result for both. Coding is ranked on the Vibe Coding Index; every other board is shown as published.
Decision reading
Neither model has a coding score yet, so the two cannot be ranked on coding.
Qwen2.5 72B Instruct has the edge where output tokens cost 2.0x less, and where what matters is Terminal-Bench Hard (Artificial Analysis run) or WeirdML, where it leads.
- Shared results
- 7
- Qwen2.5 72B Instruct only
- 46
- Qwen3 Coder Next only
- 4
- Like-for-like categories
- 5 of 9
Shared results can support a head-to-head reading. Results only one model has describe coverage, not superiority.
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
Qwen3 Coder Next
Qwen3 Coder Next leads, 9.2 against 7.7.
Same board, both measured
Agentic work
AA Agentic Index
No call
Only Qwen3 Coder Next has a AA Agentic Index result, 0.9.
Directional only
Long documents
Listed context window
Qwen3 Coder Next
Qwen3 Coder Next takes 262K tokens against 33K.
Listed by OpenRouter
Everyday cost
1K input + 500 output tokens
Qwen3 Coder Next
Qwen2.5 72B Instruct $0.0006, Qwen3 Coder Next $0.0005 per turn.
List prices
At a glance
- Coding index
- Qwen2.5 72B Instruct not benchmarked, Qwen3 Coder Next not benchmarked.
- Widest category
- Agentic work, 45.0 points apart on average over 1 shared benchmark.
- Widest single gap
- τ²-Bench Telecom: Qwen2.5 72B Instruct 34.5% against Qwen3 Coder Next 79.5%, 45.0 points apart on a 0 to 100 scale.
- Output price
- Qwen2.5 72B Instruct $0.40 against Qwen3 Coder Next $0.80 per million output tokens, 2.0x apart.
- Context window
- Qwen2.5 72B Instruct takes 33K tokens, Qwen3 Coder Next 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 72B Instruct | Bars | Qwen3 Coder Next | Ties |
|---|---|---|---|---|
| Coding | 0 | 2 | ||
| Intelligence | 0 | 1 | ||
| Agentic work | 0 | 1 | ||
| Reasoning | 0 | 2 | ||
| General | 1 | 0 | ||
| All shared boards | 1 | 6 |
CodingLike-for-likeSix of these boards feed the Vibe Coding Index02
| Benchmark | Qwen2.5 72B Instruct | Qwen3 Coder Next | Leader |
|---|---|---|---|
| AA Coding Index | Not measured | 36.2 | Only Qwen3 Coder Next measured |
| LiveCodeBench (Artificial Analysis run) | 27.6% | Not measured | Only Qwen2.5 72B Instruct measured |
| SciCode | Not measured | 36.2% | Only Qwen3 Coder Next measured |
| Terminal-Bench Hard (Artificial Analysis run) | 4.5% | 18.2% | Qwen3 Coder Next |
| WeirdML | 16.0% | 34.4% | Qwen3 Coder Next |
| LMArena Text: coding | 1292 | Not measured | Only Qwen2.5 72B Instruct measured |
| LMArena Text: software and IT | 1292 | Not measured | Only Qwen2.5 72B Instruct measured |
| Neither model has a result on the other 31 boards here | |||
IntelligenceLike-for-likePublic boards, shown as published, kept out of the index01
| Benchmark | Qwen2.5 72B Instruct | Qwen3 Coder Next | Leader |
|---|---|---|---|
| AA Intelligence Index | 7.7 | 9.2 | Qwen3 Coder Next |
| Epoch Capabilities Index | 129.0 | Not measured | Only Qwen2.5 72B Instruct measured |
Agentic workLike-for-likePublic boards, shown as published, kept out of the index01
| Benchmark | Qwen2.5 72B Instruct | Qwen3 Coder Next | Leader |
|---|---|---|---|
| AA Agentic Index | Not measured | 0.9 | Only Qwen3 Coder Next measured |
| τ²-Bench Telecom | 34.5% | 79.5% | Qwen3 Coder Next |
| METR time horizons | 35.8% | Not measured | Only Qwen2.5 72B Instruct measured |
| BALROG | 16.2% | Not measured | Only Qwen2.5 72B Instruct measured |
| TheAgentCompany | 5.7% | Not measured | Only Qwen2.5 72B Instruct measured |
| Neither model has a result on the other 34 boards here | |||
ReasoningLike-for-likePublic boards, shown as published, kept out of the index02
| Benchmark | Qwen2.5 72B Instruct | Qwen3 Coder Next | Leader |
|---|---|---|---|
| GPQA Diamond (Artificial Analysis run) | 49.1% | 73.7% | Qwen3 Coder Next |
| Humanity's Last Exam (Artificial Analysis run) | 3.6% | 10.1% | Qwen3 Coder Next |
| GPQA Diamond (Epoch AI run) | 49.1% | Not measured | Only Qwen2.5 72B Instruct measured |
| BIG-Bench Hard | 79.8% | Not measured | Only Qwen2.5 72B Instruct measured |
| LMArena Text: hard prompts | 1271 | Not measured | Only Qwen2.5 72B Instruct measured |
| LMArena Text: expert | 1245 | Not measured | Only Qwen2.5 72B Instruct measured |
| LMArena Text: hard prompts in English | 1279 | Not measured | Only Qwen2.5 72B Instruct measured |
| Neither model has a result on the other 14 boards here | |||
MathDirectional onlyPublic boards, shown as published, kept out of the index6 boards, none shared
| Benchmark | Qwen2.5 72B Instruct | Qwen3 Coder Next | Leader |
|---|---|---|---|
| AA Math Index | 14.0 | Not measured | Only Qwen2.5 72B Instruct measured |
| AIME 2025 (Artificial Analysis run) | 14.0% | Not measured | Only Qwen2.5 72B Instruct measured |
| OTIS Mock AIME | 8.1% | Not measured | Only Qwen2.5 72B Instruct measured |
| MATH level 5 | 63.2% | Not measured | Only Qwen2.5 72B Instruct measured |
| LMArena Text: math | 1283 | Not measured | Only Qwen2.5 72B Instruct measured |
| LMArena Text: mathematical work | 1290 | Not measured | Only Qwen2.5 72B Instruct measured |
| Neither model has a result on the other 7 boards here | |||
KnowledgeDirectional onlyPublic boards, shown as published, kept out of the index7 boards, none shared
| Benchmark | Qwen2.5 72B Instruct | Qwen3 Coder Next | Leader |
|---|---|---|---|
| MMLU Pro (Artificial Analysis run) | 72.0% | Not measured | Only Qwen2.5 72B Instruct measured |
| MMLU | 85.3% | Not measured | Only Qwen2.5 72B Instruct measured |
| AI2 Reasoning Challenge | 94.5% | Not measured | Only Qwen2.5 72B Instruct measured |
| LMArena Text: science | 1275 | Not measured | Only Qwen2.5 72B Instruct measured |
| LMArena Text: medicine and healthcare | 1258 | Not measured | Only Qwen2.5 72B Instruct measured |
| LMArena Text: legal and government | 1288 | Not measured | Only Qwen2.5 72B Instruct measured |
| LMArena Text: business and finance | 1253 | Not measured | Only Qwen2.5 72B Instruct measured |
| Neither model has a result on the other 10 boards here | |||
Web designNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
| Benchmark | Qwen2.5 72B Instruct | Qwen3 Coder Next | Leader |
|---|---|---|---|
| Neither model has a result on 9 boards here | |||
GeneralLike-for-likePublic boards, shown as published, kept out of the index10
| Benchmark | Qwen2.5 72B Instruct | Qwen3 Coder Next | Leader |
|---|---|---|---|
| IFBench | 36.9% | 35.2% | Qwen2.5 72B Instruct |
| Long context reasoning (AA-LCR) | Not measured | 47.0% | Only Qwen3 Coder Next measured |
| DTBench | 62.9% | Not measured | Only Qwen2.5 72B Instruct measured |
| LMCA | 13.4% | Not measured | Only Qwen2.5 72B Instruct measured |
| HellaSwag | 84.8% | Not measured | Only Qwen2.5 72B Instruct measured |
| WinoGrande | 82.3% | Not measured | Only Qwen2.5 72B Instruct measured |
| PIQA | 82.6% | Not measured | Only Qwen2.5 72B Instruct measured |
| LMArena Text | 1269 | Not measured | Only Qwen2.5 72B Instruct measured |
| LMArena Text: instruction following | 1254 | Not measured | Only Qwen2.5 72B Instruct measured |
| LMArena Text: creative writing | 1221 | Not measured | Only Qwen2.5 72B Instruct measured |
| LMArena Text: English | 1283 | Not measured | Only Qwen2.5 72B Instruct measured |
| LMArena Text: non-English | 1252 | Not measured | Only Qwen2.5 72B Instruct measured |
| LMArena Text: Chinese | 1273 | Not measured | Only Qwen2.5 72B Instruct measured |
| LMArena Text: German | 1234 | Not measured | Only Qwen2.5 72B Instruct measured |
| LMArena Text: Japanese | 1180 | Not measured | Only Qwen2.5 72B Instruct measured |
| LMArena Text: Korean | 1188 | Not measured | Only Qwen2.5 72B Instruct measured |
| LMArena Text: French | 1280 | Not measured | Only Qwen2.5 72B Instruct measured |
| LMArena Text: Spanish | 1255 | Not measured | Only Qwen2.5 72B Instruct measured |
| LMArena Text: Russian | 1264 | Not measured | Only Qwen2.5 72B Instruct measured |
| LMArena Text: multi-turn | 1272 | Not measured | Only Qwen2.5 72B Instruct measured |
| LMArena Text: longer queries | 1281 | Not measured | Only Qwen2.5 72B Instruct measured |
| LMArena Text: writing and language | 1247 | Not measured | Only Qwen2.5 72B Instruct measured |
| LMArena Text: entertainment and media | 1211 | Not measured | Only Qwen2.5 72B Instruct measured |
| Neither model has a result on the other 23 boards here | |||
Image generationNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
| Benchmark | Qwen2.5 72B Instruct | Qwen3 Coder Next | 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
Qwen3 Coder Next costs less
Repository review
50K input + 3K output tokens
Qwen3 Coder Next costs less
Agent loop
200K input + 10K output tokens
Qwen3 Coder Next 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 72B Instruct | Qwen3 Coder Next |
|---|---|---|
| Maker | Qwen | Qwen |
| Weights | Open weights | Open weights |
| Released | September 19, 2024 | February 4, 2026 |
| Context window | 33K | 262K |
| Input $/Mtok | $0.36 | $0.12 |
| Output $/Mtok | $0.40 | $0.80 |
| Coding index | Not benchmarked | Not benchmarked |
| Independent measurements behind it | 0 | 0 |
| Cost per solved task | Not computable | Not computable |
| Source | OpenRouter listing | OpenRouter listing |
Questions
Is Qwen2.5 72B Instruct better than Qwen3 Coder Next for coding?
Neither model has a coding score yet, so the two cannot be ranked on coding. Qwen2.5 72B Instruct has the edge where output tokens cost 2.0x less, and where what matters is Terminal-Bench Hard (Artificial Analysis run) or WeirdML, where it leads.
Which is cheaper, Qwen2.5 72B Instruct or Qwen3 Coder Next?
Qwen2.5 72B Instruct costs $0.36 per million input tokens and $0.40 per million output tokens; Qwen3 Coder Next costs $0.12 and $0.80. Cost per solved task is known for neither of them.
Which has the bigger context window, Qwen2.5 72B Instruct or Qwen3 Coder Next?
Qwen3 Coder Next, with 262K tokens against 33K.
How many benchmarks is the Qwen2.5 72B Instruct vs Qwen3 Coder Next comparison based on?
7 benchmarks have a result for both models as of October 21, 2026. Qwen2.5 72B Instruct's coding index rests on 0 independent benchmarks and Qwen3 Coder Next's on 0. Where only one model has a result, the page says so instead of naming a winner.