DeepSeek V4 Pro vs Qwen3.8-Flash-Next
Full profiles: DeepSeek V4 Pro (DeepSeek) and Qwen3.8-Flash-Next (Alibaba): every benchmark, price and rank.
Updated · Data from 8 sources
As of 16 Nov 2026, DeepSeek V4 Pro and Qwen3.8-Flash-Next are neck and neck on quality (56.0 vs 53.9 on the BenchLeader Index, inside the ±9.0 margin), and Qwen3.8-Flash-Next is 4.3× cheaper at $0.230 vs $0.990 per 1M tokens blended. Choose DeepSeek V4 Pro for reasoning (56 vs 43) and knowledge (41 vs 38) and fast first answers (42.8 s vs 65.5 s); choose Qwen3.8-Flash-Next for coding (61 vs 52), where it also streams ~1.5× faster.
Key facts (as of 16 Nov 2026)
- DeepSeek V4 Pro scores 56.0 on the BenchLeader Index and Qwen3.8-Flash-Next scores 53.9 (BenchLeader, 16 Nov 2026).
- DeepSeek V4 Pro costs $0.990 per 1M tokens blended (list price, 16 Nov 2026).
- Qwen3.8-Flash-Next costs $0.230 per 1M tokens blended (list price, 16 Nov 2026).
- Qwen3.8-Flash-Next streams ~1.5× faster than DeepSeek V4 Pro (Artificial Analysis, 7-day median, 16 Nov 2026).
- Of the 18 benchmarks both models have, Qwen3.8-Flash-Next leads 10 (BenchLeader, 16 Nov 2026).
- On Arena WebDev, DeepSeek V4 Pro is rated 1581 and Qwen3.8-Flash-Next 1624 (LMArena, as published by 16 Nov 2026).
Which to pick for…
The pick's numbers come first. Category scores on the BenchLeader scale (50 is the average model, 15 points one standard deviation). High confidence means a lead of half a standard deviation or more; under one point is called either.
| Use case | Pick | Why | Confidence |
|---|---|---|---|
| Coding | Qwen3.8-Flash-Next | 61 vs 52 category score; Arena WebDev 1624 vs 1581 | High |
| Reasoning | DeepSeek V4 Pro | 56 vs 43 category score; HLE-Diamond (AA) 20.5% vs 16.3% | High |
| Knowledge | DeepSeek V4 Pro | 41 vs 38 category score; AA-Omniscience v1.1 -6.5 vs -12.2 | Medium |
| Long documents | Either | 64 vs 65 category score; AA-LCR v1.1 79.3% vs 79.7% | Either |
| Chat quality (human preference) | – | Only DeepSeek V4 Pro has an Arena rating (65 category score) | Not comparable |
| Lowest cost | Qwen3.8-Flash-Next | $0.230 vs $1.98 per 1M blended; agentic coding session $0.011 vs $0.095 | – |
| Fastest first answer | DeepSeek V4 Pro | 42.8 s (DeepSeek V4 Pro (high)) vs 65.5 s (Qwen3.8-Flash-Next) to the first answer token | – |
| High volume on a budget | Neither flagship | A smaller model is usually the better fit: Cheapest good LLMs, Fastest LLMs | – |
Key differences
- DeepSeek V4 Pro (high) and Qwen3.8-Flash-Next are neck and neck on quality (56.0 vs 53.9, within the ±9.0 margin).
- DeepSeek V4 Pro (high) is stronger in knowledge, reasoning.
- Qwen3.8-Flash-Next is stronger in coding, long context.
- Qwen3.8-Flash-Next is 4.3× cheaper ($0.230 vs $0.990 per 1M blended).
At a glance
| Spec | DeepSeek V4 Pro | Qwen3.8-Flash-Next |
|---|---|---|
| Developer | DeepSeek | Alibaba |
| Released | 13 Aug 2026 | 26 Aug 2026 |
| Context window (tokens) | 1,048,576 | 256,000 |
| Compared at | DeepSeek V4 Pro (high) | Qwen3.8-Flash-Next |
| Open weights | Yes | Yes |
| Price in / out / cached input per 1M | $1.32 / $3.96 / $0.022 | $0.150 / $0.470 / – |
Category by category
Category scores on the BenchLeader scale: 50 is the average evaluated model, 15 points is one standard deviation. A gap under one point is called level.
| Category | DeepSeek V4 Pro (high) | Qwen3.8-Flash-Next | Leader |
|---|---|---|---|
| Reasoning | 55.6 | 43.0 | DeepSeek V4 Pro |
| Coding | 51.5 | 60.7 | Qwen3.8-Flash-Next |
| Agents & tools | – | 64.0 | not comparable |
| Maths | 64.5 | – | not comparable |
| Knowledge | 41.2 | 38.1 | DeepSeek V4 Pro |
| Human preference | 65.0 | – | not comparable |
| Multimodal | – | 61.0 | not comparable |
| Long context | 64.3 | 64.5 | level |
On one chart
Each domain on the index's own scale, where 50 is an average model. Click a domain to see the benchmarks behind it. Pick up to 5 models from the list.
Not drawn, measured for fewer than two of these models: Agents & tools, Maths, Instruction following, Multimodal, Human preference.
Where sources disagree
Across the 18 benchmarks both have, DeepSeek V4 Pro leads 8, Qwen3.8-Flash-Next leads 10. Every overall index that rates both puts Qwen3.8-Flash-Next ahead.
| Overall index | DeepSeek V4 Pro (high) | Qwen3.8-Flash-Next | Leader |
|---|---|---|---|
| BenchLeader Index | 56.0 | 53.9 | neck and neck |
| AA Intelligence Index v5.0.1 | 20.8 | 28.6 | Qwen3.8-Flash-Next |
Each publisher runs its own tasks, harnesses and effort settings, so their overall verdicts can differ; BenchLeader reconciles them on one scale. † measured at another effort setting of the same model (the publisher did not test the one compared here).
Price, speed and latency
List prices blended at three input tokens per output token. Per-task prices use the token counts on the methodology page. Cost, tokens and time per benchmark task are measured: Artificial Analysis’ averages over its whole Intelligence Index run.
| Measure | DeepSeek V4 Pro (high) | Qwen3.8-Flash-Next |
|---|---|---|
| Blended price per 1M tokens | $1.98 | $0.230 |
| Input / output per 1M | $1.32 / $3.96 | $0.150 / $0.470 |
| Cached input per 1M | $0.022 | – |
| Price per task: chat reply | $0.0017 | $0.0002 |
| Price per task: summarize a 30-page report | $0.018 | $0.0021 |
| Price per task: code edit | $0.014 | $0.0016 |
| Price per task: agentic coding session | $0.095 | $0.011 |
| Price per task: structured extraction | $0.0034 | $0.0004 |
| Agentic coding session, input mostly cached | $0.037 | – |
| Cost per benchmark task (AA Index) | $0.849 | $0.569 |
| Output tokens per task (AA Index) | 70k | 148k |
| Of which reasoning | 45k | 97k |
| Time per task (AA Index) | 15 min | 31 min |
| Output speed | 80 tok/s | 80 tok/s |
| Time to first answer | 42.8 s | 65.5 s |
| Time to a complete answer, last measured 13 Nov 2026 | – | 202.8 s |
Per token is not per task: an average task of Artificial Analysis' Intelligence Index cost $0.849 on DeepSeek V4 Pro (high) and $0.569 on Qwen3.8-Flash-Next, and Qwen3.8-Flash-Next wrote 2.1× as many output tokens for the same work, which a per-token price hides.
Each model is compared at its best-ranked reasoning-effort setting. DeepSeek V4 Pro (high) scores 56.0 and takes 43 s to its first answer token; DeepSeek V4 Pro (max) scores 54.7 in 47 s, DeepSeek V4 Pro scores 53.0 in 1 s. Qwen3.8-Flash-Next scores 53.9 and takes 66 s to its first answer token. Higher effort means more thinking before the answer: better scores, longer waits and more tokens billed. How reasoning effort changes scores and cost.
Every benchmark
127 benchmarks, grouped by category; each name links to the board and its leaders. † measured at another effort setting of the same model, shown for reference and not counted as a win.
| Metric | Publisher | DeepSeek V4 Pro (high) | Qwen3.8-Flash-Next |
|---|---|---|---|
| BenchLeader Index | BenchLeader | 56.0 | 53.9 |
| Coding score | BenchLeader | 51.5 | 60.7 |
| Human preference score | BenchLeader | 65.0 | – |
| Knowledge score | BenchLeader | 41.2 | 38.1 |
| Long context score | BenchLeader | 64.3 | 64.5 |
| Maths score | BenchLeader | 64.5 | – |
| Reasoning score | BenchLeader | 55.6 | 43.0 |
| Agents & tools score | BenchLeader | – | 64.0 |
| Multimodal score | BenchLeader | – | 61.0 |
| Blended price $/M | BenchLeader | $1.98 | $0.230 |
| Output speed | BenchLeader | 80 tok/s | 80 tok/s |
| Time to first answer | BenchLeader | 42.8 s | 65.5 s |
| Context window | BenchLeader | 1,048,576 tokens | 256,000 tokens |
| Reasoning | |||
| GPQA Diamond | Epoch AI Benchmarking Hub | 90.9% | – |
| Arena Hard Prompts | Arena (LMArena) | 1482 | – |
| LiveBench 1.0 Reasoning | LiveBench | 85.8%† | 87.4% |
| GPQA Diamond (AA) | Artificial Analysis | 92.8%† | 92.3% |
| Humanity's Last Exam (AA) | Artificial Analysis | 41.5% | 38.0% |
| HLE-Diamond (AA) | Artificial Analysis | 20.5% | 16.3% |
| GPQA Diamond (Vals) | Vals AI | 92.4%† | – |
| Kagi LLM Benchmark | Kagi LLM Benchmark | 53.5%† | – |
| ARC-AGI-1 | ARC Prize | 87.2% | – |
| ARC-AGI-2 | ARC Prize | 59.7% | – |
| CritPt | Artificial Analysis | 18.6% | 11.1% |
| CritPt v2 hard subset (AA) | Artificial Analysis | 28.5% | 24.5% |
| HLE-Diamond (Scale) | Scale AI SEAL | 13.4%† | – |
| Arena Expert | Arena (LMArena) | 1481 | – |
| Chess Puzzles | Epoch AI Benchmarking Hub | 16.0% | – |
| Mystery Game Puzzles | Epoch AI Benchmarking Hub | 43.0%† | – |
| CL-bench Life | Epoch AI Benchmarking Hub | 13.5% | – |
| LMCA | Epoch AI Benchmarking Hub | 45.5%† | – |
| DTBench | Epoch AI Benchmarking Hub | 93.9%† | – |
| ForecastBench | ForecastBench | 60.7%† | – |
| Coding | |||
| SWE-bench Verified (Epoch) | Epoch AI Benchmarking Hub | 77.6%† | – |
| SciCode | SciCode | 46.4% | – |
| WeirdML | WeirdML | 46.5% | – |
| FrontierCode | Cognition | 28.6% | – |
| Arena Coding | Arena (LMArena) | 1503 | – |
Show all 127 benchmarks
| Metric | Publisher | DeepSeek V4 Pro (high) | Qwen3.8-Flash-Next |
|---|---|---|---|
| Coding | |||
| Arena WebDev | Arena (LMArena) | 1581 | 1624 |
| LiveBench 1.0 Coding | LiveBench | 77.2%† | 72.5% |
| SciCode (AA) | Artificial Analysis | 51.7% | 50.6% |
| LiveCodeBench | Vals AI | 87.5%† | – |
| IOI | Vals AI | 51.6%† | – |
| SWE-bench (Vals) | Vals AI | 96.4%† | – |
| Code Migration | Vals AI | 41.5%† | – |
| Code Migration v1.1 | Vals AI | 52.0%† | – |
| ProgramBench | Vals AI | 0.0%† | – |
| Vibe Code Bench 1-100 | Vals AI | 17.5%† | – |
| Vibe Code Bench v1.1 | Vals AI | 82.3%† | – |
| SWE Atlas v1.0: Codebase QnA | Scale AI SEAL | 27.1%† | – |
| SWE Atlas v1.0: Test Writing | Scale AI SEAL | 27.1%† | – |
| MirrorCode | Epoch AI Benchmarking Hub | 33.3%† | – |
| Surface Evolver Bench | Epoch AI Benchmarking Hub | 40.0% | – |
| ALE-Bench | Epoch AI Benchmarking Hub | 1006.1 | – |
| Terminal-Bench 4.0 (AA) | Artificial Analysis | 6.1% | 25.3% |
| Terminal-Bench 2.1 (AA) | Artificial Analysis | 78.7%† | 86.1% |
| Startrise Frontend | Startrise | 62.7%† | – |
| GameBench 2 Overall | DuelLab GameBench 2 | 41.9† | – |
| GameBench 2 by reasoning setting | DuelLab GameBench 2 | 45.1 | – |
| Design Arena Website | Design Arena | 1227† | – |
| FrontierCode 1.1 (AA) | Artificial Analysis | 38.8%† | – |
| FrontierCode 2.0 (AA) | Artificial Analysis | 16.7% | 27.1% |
| Agents & tools | |||
| Terminal-Bench | Terminal-Bench | 6.7%† | – |
| APEX-Agents | Mercor | 47.3%† | – |
| Arena Agent | Arena (LMArena) | – | -4.6 |
| LiveBench Agentic Coding | LiveBench | 55.0%† | 61.6% |
| Terminal-Bench 2.1 (Vals) | Vals AI | 50.2%† | – |
| HiL-Bench | Scale AI SEAL | 38.5%† | – |
| GDPval-AA v2.1, normalised score | Artificial Analysis | 47.5%† | 56.4% |
| GDPval-AA v2.1 | Artificial Analysis | 1336 | 1637 |
| τ³-Banking (AA) | Artificial Analysis | 39.6%† | 45.4% |
| CyberBench | Vals AI | 72.9%† | – |
| Finance Agent v2 | Vals AI | 50.4%† | – |
| Finance Agent v2.1 | Vals AI | 57.1%† | – |
| Harvey's Legal Agent Benchmark v1.1 | Vals AI | 10.8%† | – |
| Harvey's Legal Agent Benchmark v1 | Vals AI | 7.5%† | – |
| Legal Research Bench | Vals AI | 40.9%† | – |
| SkillsBench | Vals AI | 53.8%† | – |
| SREBench | Vals AI | 1.9%† | – |
| Tax Agent Bench | Vals AI | 26.8%† | – |
| Terminal-Bench 2.0 (Vals) | Vals AI | 56.2%† | – |
| Terminal-Bench 4.0 (Vals) | Vals AI | 14.1%† | – |
| Terminal-Bench Science | Vals AI | 4.3%† | – |
| Arena Agent Task Completion | Arena (LMArena) | – | 0.3 |
| Vending-Bench 2 | Andon Labs | 3284.5† | – |
| AutomationBench | Artificial Analysis | 46.2% | 55.9% |
| GDP.pdf (AA) | Artificial Analysis | 10.4% | 15.6% |
| EnterpriseOps-Gym | Artificial Analysis | 49.6%† | – |
| AA-Briefcase v1.1 | Artificial Analysis | 1148 | 1572 |
| Design Arena Agentic | Design Arena | 927† | – |
| Terminal-Bench-Science 0.1 (AA) | Artificial Analysis | 1.9% | 7.1% |
| Maths | |||
| FrontierMath Tiers 1–3 | Epoch AI Benchmarking Hub | 64.6%† | – |
| FrontierMath Tier 4 | Epoch AI Benchmarking Hub | 26.8%† | – |
| OTIS Mock AIME | Epoch AI Benchmarking Hub | 95.6% | – |
| ProofBench | Vals AI | 50.0%† | – |
| LiveBench Mathematics | LiveBench | 95.1%† | 85.8% |
| AIME 2026 | MathArena | 96.7%† | – |
| HMMT February 2026 | MathArena | 93.9%† | – |
| USAMO 2026 | MathArena | 60.7%† | – |
| MathArena Apex | MathArena | 28.1%† | – |
| Arena Maths | Arena (LMArena) | 1465 | – |
| Knowledge | |||
| SimpleQA Verified | Epoch AI Benchmarking Hub | 52.9%† | – |
| LiveBench 1.0 Data Analysis | LiveBench | 79.2%† | 74.2% |
| AA-Omniscience v1.1 | Artificial Analysis | -6.5 | -12.2 |
| MMLU-Pro | Vals AI | 87.3%† | – |
| LegalBench | Vals AI | 82.4%† | – |
| CorpFin | Vals AI | 65.4%† | – |
| TaxEval | Vals AI | 73.1%† | – |
| CaseLaw v2 | Vals AI | 59.4%† | – |
| Excel Modeling Benchmark | Vals AI | 52.8%† | – |
| Excel Modeling Benchmark v1.1 | Vals AI | 65.1%† | – |
| MedCode | Vals AI | 42.5%† | – |
| MedScribe | Vals AI | 80.2%† | – |
| Public Benefits Bench v1.1 | Vals AI | 62.9%† | – |
| Immigration Bench | Vals AI | 61.7%† | – |
| DrugDiscoveryBench | Scale AI SEAL | 31.7%† | – |
| AA-Omniscience v1.1: accuracy | Artificial Analysis | 45.1% | 23.5% |
| AA-Omniscience v1.1: non-hallucination | Artificial Analysis | 5.9% | 53.2% |
| Instruction following | |||
| LiveBench Language | LiveBench | 82.1%† | 74.6% |
| LiveBench Instruction Following | LiveBench | 67.7%† | 77.1% |
| Arena Instruction Following | Arena (LMArena) | 1461 | – |
| Human preference | |||
| Arena Text | Arena (LMArena) | 1464 | – |
| EQ-Bench 4 | EQ-Bench | 1156† | – |
| EQ-Bench Creative Writing v3 | EQ-Bench | 1553† | – |
| Arena Creative Writing | Arena (LMArena) | 1444 | – |
| Arena Multi-turn | Arena (LMArena) | 1469 | – |
| Arena Longer Queries | Arena (LMArena) | 1474 | – |
| Multimodal | |||
| MMMU-Pro | Artificial Analysis | – | 78.6% |
| Long context | |||
| AA-LCR v1.1 | Artificial Analysis | 79.3% | 79.7% |
| MLCR | Artificial Analysis | – | 20.6% |
| Composite | |||
| Epoch Capabilities Index | Epoch AI Benchmarking Hub | 155.9† | – |
| LiveBench 1.0 | LiveBench | 77.4%† | 76.2% |
| AA Intelligence Index v5.0.1 | Artificial Analysis | 20.8 | 28.6 |
| AA Intelligence Index v4.3.2 | Artificial Analysis | 36† | 39.8 |
| AA Progress Index v1 | Artificial Analysis | 311.5† | 320.2 |
| Vals Index | Vals AI | 53.0† | – |
| Vals Finance Index | Vals AI | 52.9%† | – |
| Vals Router Index | Vals AI | 53.0%† | – |
| Vals Legal Index | Vals AI | 69.0%† | – |
| Metadata | |||
| AA Openness Index | Artificial Analysis | 44.4 | 38.9 |
Data as of 2026-11-16. Best configuration of each model; every score links to its source on the model pages.
DuelLab GameBench 2, release gb2-20261111-7b598bb13de4 (https://benchmarks.duellab.org/releases/gb2-20261111-7b598bb13de4/), CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). Scores mapped to BenchLeader model names; not endorsed by DuelLab.
What changed
- DeepSeek V4 Pro (high): BenchLeader Index 59.1 → 56.0.
- DeepSeek V4 Pro (high): 17 new results (AA Intelligence Index v5.0.1, Humanity's Last Exam (AA), HLE-Diamond (AA)…); input price moved from $0.66/M to $1.32/M (100% up); output price moved from $1.98/M to $3.96/M (100% up).
- Qwen3.8-Flash-Next: BenchLeader Index 54.7 → 54.0.
- Qwen3.8-Flash-Next: 3 new results (AA Intelligence Index v5.0.1, FrontierCode 2.0 (AA), Terminal-Bench-Science 0.1 (AA)).
- Qwen3.8-Flash-Next: 1 new result (CritPt v2 hard subset (AA)); 1 score revision.
- DeepSeek V4 Pro (high): BenchLeader Index 57.5 → 59.1.
- DeepSeek V4 Pro (high): 1 score revision.
- DeepSeek V4 Pro (high): 1 score revision.
- Qwen3.8-Flash-Next: 1 new result (AA-Briefcase v1.1); 1 score revision.
- Qwen3.8-Flash-Next: BenchLeader Index 58.7 → 54.7.
Related comparisons
- DeepSeek V4 Pro vs Claude Fable 5.2
- DeepSeek V4 Pro vs Claude Opus 5.5
- Qwen3.8-Flash-Next vs Claude Opus 5.5
- Qwen3.8-Flash-Next vs GPT-6 Astra
- Qwen3.8-Flash-Next vs GPT-6.1 Sol
- Qwen3.8-Flash-Next vs Claude Fable 5
- DeepSeek V4 Pro vs Claude Fable 5.1
- DeepSeek V4 Pro vs GPT-6.1 Sol
Rankings: Best LLM overall · Best LLM for agents · Best cheap LLMs
DeepSeek V4 Pro vs Qwen3.8-Flash-Next: questions
- Is DeepSeek V4 Pro better than Qwen3.8-Flash-Next?
- DeepSeek V4 Pro (high) and Qwen3.8-Flash-Next are neck and neck on quality (56.0 vs 53.9, within the ±9.0 margin). The BenchLeader Index combines every independent quality benchmark; the two are within the margin of error as of 16 Nov 2026, so check the category scores for your use.
- DeepSeek V4 Pro vs Qwen3.8-Flash-Next: which is cheaper, and will that change?
- Qwen3.8-Flash-Next is cheaper today: $0.230 against $0.990 per million tokens, blended at three input tokens per output token. No price change has been announced for either.
- Is DeepSeek V4 Pro or Qwen3.8-Flash-Next better for coding?
- Qwen3.8-Flash-Next scores higher in coding (61 vs 52 on the category index, where 50 is average). On Arena WebDev, the top shared coding benchmark, it is 1581 for DeepSeek V4 Pro against 1624 for Qwen3.8-Flash-Next.
- Which is faster, DeepSeek V4 Pro or Qwen3.8-Flash-Next?
- Qwen3.8-Flash-Next streams ~1.5× faster (80 vs 80 output tokens per second, Artificial Analysis 7-day median).
- Which is better for long documents?
- They are level on long-context benchmarks (64 vs 65). DeepSeek V4 Pro accepts more context: 1,048,576 against 256,000 tokens for Qwen3.8-Flash-Next.
More questions (3)
- Which has lower latency?
- DeepSeek V4 Pro answers first: 42.8 s against 65.5 s to the first answer token, as Artificial Analysis measures it. Reasoning effort drives most of this: a higher setting thinks longer before it answers.
- Do independent benchmarks agree on DeepSeek V4 Pro vs Qwen3.8-Flash-Next?
- Across the 18 benchmarks both have, DeepSeek V4 Pro leads 8, Qwen3.8-Flash-Next leads 10. Every overall index that rates both puts Qwen3.8-Flash-Next ahead.
- When was Qwen3.8-Flash-Next released?
- Qwen3.8-Flash-Next was released on 26 Aug 2026.