Across 34 comparable benchmarks, DeepSeek-V4.1-Flash has better reported results on 28 and worse on 6. 3 further shared results are listed below, each with the reason it is not counted.
Moderate evidence34 of 37 shared results countedHow we compare
DeepSeek-V4.1-Flash is no longer available through its official API. API prices for this version are not shown.
Price basis: Listed rates vs listed rates. The price section below shows the applicable periods.
| Fact | DeepSeek-V4.1-Flash | Qwen3.8-27B |
|---|---|---|
| Context window | 1M tokens | 256K tokens |
| Input / 1M | — | $0.500 |
| Output / 1M | — | $3.00 |
| Blended (4:1) | — | $0.712 |
| Status | Retired | Available |
| Released | September 10, 2026 | August 14, 2026 |
Blended is 4:1 input to output with 90% cache hits at listed rates — set your own mix.
Mean of 20 shared benchmarks grouped by capability on a 0–100 scale. How averages are built
DeepSeek-V4.1-Flash ahead on 9 of 9 counted
DeepSeek-V4.1-Flash ahead on 2 of 2 counted
DeepSeek-V4.1-Flash ahead on 1 of 1 counted
DeepSeek-V4.1-Flash ahead on 5 of 5 counted
Qwen3.8-27B ahead on 1 of 1 counted
No fixed-scale average — see the rows in Benchmark results
DeepSeek-V4.1-Flash ahead on 4 of 5 counted
Qwen3.8-27B ahead on 4 of 7 counted
No shared comparable results for Research & search, Long context, Cybersecurity.
Published scores, with matching modes where available. Gaps show DeepSeek-V4.1-Flash relative to Qwen3.8-27B. Reading the results
| Benchmark | DeepSeek-V4.1-Flash | Qwen3.8-27B | Gap |
|---|---|---|---|
| AA Cyber IndexRepository Engineering · Composite index (0–100) | 40.8Max · tools | 12.7≠Extra-High · tools | 3.21×#11 vs #21 of 22 |
| APEX-SWERepository Engineering · Pass@1 | 51.9Max · tools | 50.7≠Extra-High · tools | +1.2 pts#12 vs #13 of 61 |
| Code MigrationMaintenance & Optimization · Hidden-test pass rate (%) | 45.6Thinking High · tools | 14.2≠Extra-High · tools | +31.5 pts#18 vs #59 of 79 |
| DeepSWERepository Engineering · Pass@1 (DeepSWE v1.1) | 74.2Max · tools | 42.2≠Thinking · tools | +32.0 pts#11 vs #64 of 73 |
| FrontierCode 2.0Repository Engineering · Weighted rubric score | 19.3Max · tools + web | 14.9≠Thinking · tools | 1.30×#22 vs #27 of 36 |
| IOI (Vals v2)Algorithmic Coding · Subtask points (%) | 40.3Thinking High · tools | 39.1≠Extra-High · tools | +1.2 pts#35 vs #37 of 45 |
| NL2Repo-BenchRepository Engineering · Average test pass rate | 65.4Max · tools | 42.3≠Thinking · tools | +23.1 pts#3 vs #18 of 21 |
| Terminal-Bench 2.1Agentic Development · Accuracy | 90.6Max · tools | 73.0≠Thinking · tools | +17.6 pts#5 vs #46 of 68 |
| Terminal-Bench 4.0Agentic Development · Resolution rate (%)Not counted · Different or unrecorded test environments — excluded from the tally | 35.4Max · tools· vals-effort-study | 5.6≠Extra-High · tools· legacy-default | —#33 vs #67 of 90 |
| Vibe Code Bench v1.1Code Generation & Editing · Pass rate (%) | 84.7Thinking High · tools | 64.8≠Extra-High · tools | +19.9 pts#23 vs #54 of 121 |
| ARC-AGI-1Abstract Generalization · Score (%) | 94.5Max · no tools | 87.5≠Extra-High · no tools | +7.0 pts#21 vs #40 of 87 |
| ARC-AGI-2Abstract Generalization · Score (% solved); cost per task (USD)Not counted · Speed and cost telemetry is reported, not scored | 72.9Max · no tools | 42.4≠Extra-High · no tools | 1.72×#26 vs #48 of 80 |
| SimpleBenchCommonsense · Score (AVG@5) | 66.7Historical | 60.2≠Thinking · no tools | 1.11×#33 vs #48 of 106 |
| ProofBench v1.1 (Lean 4)Mathematics · Accuracy (%) | 54.0Thinking High · tools | 16.0≠Extra-High · tools | +38.0 pts#31 vs #46 of 54 |
| CritPtScientific Reasoning · Score | 14.3Max · no tools | 5.4≠Extra-High · no tools | 2.65×#43 vs #68 of 134 |
| GPQA DiamondScientific Reasoning · Accuracy | 90.9Max · no tools | 89.2≠Thinking · no tools | +1.7 pts#44 vs #58 of 200 |
| SciCodeScientific Computing · Score | 51.9Max · no tools | 46.6≠Extra-High · no tools | 1.11×#43 vs #58 of 98 |
| MedCodeClinical Workflows · Accuracy (%) | 41.2Thinking High · tools | 28.7≠Extra-High · tools | +12.5 pts#54 vs #91 of 97 |
| MedScribeClinical Workflows · Rubric pass rate (%) | 85.5Thinking High · tools | 83.8≠Extra-High · tools | +1.7 pts#26 vs #39 of 99 |
| APEX-AgentsTool Orchestration · Score | 39.5Reported best | 47.5Reported best | 0.83×#47 vs #37 of 61 |
| BabyVisionVisual Understanding · Accuracy | 89.6Max · tools | 85.6≠Thinking · tools | +4.0 pts#10 vs #13 of 17 |
| ChartographyDocuments & Charts · Score | 78.9Max · tools | 52.1≠Thinking · tools | 1.51×#12 vs #24 of 29 |
| GDP.pdfDocuments & Charts · Task pass rate (%) | 12.8Max · no tools | 16.4≠Extra-High · no tools | −3.6 pts#50 vs #39 of 96 |
| MMMU-ProVisual Understanding · Accuracy | 77.0Max · no tools | 76.3≠Extra-High · no tools | +0.7 pts#52 vs #58 of 136 |
| ZeroBench MainVisual Understanding · Pass@1 / Pass@5 / Pass^5 | 49.0Max · tools | 12.0≠Thinking · no tools | +37.0 pts#10 vs #19 of 21 |
| Creative WritingWriting | 1540Normal · no tools | 1671Normal · no tools | 0.92×#54 vs #38 of 121 |
| EMB (Excel Modeling)Finance · Accuracy (%) | 57.2Thinking High · tools | 59.7≠Extra-High · tools | −2.4 pts#42 vs #38 of 78 |
| Finance Agent v2Finance · Score | 53.5Thinking High · tools | 48.5≠Extra-High · tools | 1.10×#33 vs #47 of 75 |
| Harvey's Legal Agent BenchmarkLegal · Accuracy (%) | 6.7Thinking High · tools | 11.3≠Extra-High · tools | −4.6 pts#21 vs #11 of 52 |
| Legal Research BenchLegal · Score | 41.4Thinking High · tools | 36.1≠Extra-High · tools | 1.15×#32 vs #49 of 91 |
| SAGEService Workflows · Accuracy (%) | 47.9Thinking High · tools | 52.4≠Extra-High · tools | −4.5 pts#29 vs #7 of 82 |
| Tax Agent BenchFinance · Accuracy (%, rubric × citation quality) | 62.5Thinking High · tools | 30.9≠Extra-High · tools | +31.6 pts#15 vs #41 of 74 |
| HLEKnowledge Exams · Accuracy | 63.9Max · tools | 30.8≠Thinking · no tools | +33.1 pts#7 vs #86 of 151 |
| ECICapability Indices · ECI score (capability index, higher is better) | 155Reported best | 149Reported best | 1.04×#30 vs #52 of 168 |
| LiveBenchCross-capability Suites · Accuracy | 81.1Max · no tools | 75.3≠Reported best | +5.8 pts#8 vs #34 of 106 |
| LiveBench 2.0Cross-capability Suites · Overall score (equal-weight mean of 7 suite scores, 0-100) | 78.0Max · task tools | 62.7Max · task tools | 1.25×#12 vs #36 of 51 |
| Vals IndexCapability IndicesNot counted · This metric has no better direction | 57.9Thinking High · tools | 48.5≠Extra-High · tools | — |
Repository Engineering · Composite index (0–100)
Repository Engineering · Pass@1
Maintenance & Optimization · Hidden-test pass rate (%)
Repository Engineering · Pass@1 (DeepSWE v1.1)
Repository Engineering · Weighted rubric score
Algorithmic Coding · Subtask points (%)
Repository Engineering · Average test pass rate
Agentic Development · Accuracy
Agentic Development · Resolution rate (%)
Not counted · Different or unrecorded test environments — excluded from the tally
Code Generation & Editing · Pass rate (%)
Abstract Generalization · Score (%)
Abstract Generalization · Score (% solved); cost per task (USD)
Not counted · Speed and cost telemetry is reported, not scored
Commonsense · Score (AVG@5)
Mathematics · Accuracy (%)
Scientific Reasoning · Score
Scientific Reasoning · Accuracy
Scientific Computing · Score
Clinical Workflows · Accuracy (%)
Clinical Workflows · Rubric pass rate (%)
Tool Orchestration · Score
Visual Understanding · Accuracy
Documents & Charts · Score
Documents & Charts · Task pass rate (%)
Visual Understanding · Accuracy
Visual Understanding · Pass@1 / Pass@5 / Pass^5
Writing
Finance · Accuracy (%)
Finance · Score
Legal · Accuracy (%)
Legal · Score
Service Workflows · Accuracy (%)
Finance · Accuracy (%, rubric × citation quality)
Knowledge Exams · Accuracy
Capability Indices · ECI score (capability index, higher is better)
Cross-capability Suites · Accuracy
Cross-capability Suites · Overall score (equal-weight mean of 7 suite scores, 0-100)
Capability Indices
Not counted · This metric has no better direction
A dash means no result is available.
Time-of-day providers use peak-hour rates; fixed-price providers use standard rates. Off-peak discounts remain in the published details.
Drag the handle above the bar for cache hits, the one below it for the input : output mix · 5% steps
Blended price = effective input rate at 90% cache hits × 80% + output rate × 20%
A blended cost needs a published input and output price on both sides, and one of these two does not have them yet.
DeepSeek-V4.1-Flash: Listed text rates; pricing tier not confirmed.
Qwen3.8-27B: International pricing. context_length <= 1000000 Alibaba Model Studio International (Singapore)
| Price / 1M tokens | DeepSeek-V4.1-Flash | Qwen3.8-27B | Ratio |
|---|---|---|---|
| Input | — | $0.50 | — |
| Output | — | $3.00 | — |
| Cached input | — | $0.10 | — |
| Cache write | — | — | — |
| Blended | — | $0.712 | — |
Estimates use the selected token mix. Cache writes, tool fees and long-context surcharges are excluded.
Reported cost per task, paired with the score from the same run. Settings and sources may differ; costs do not affect the verdict.
DeepSeek-V4.1-Flash relative to Qwen3.8-27B
Summary: 3 of 4 shared benchmarks have matching recorded settings and sources. The average is a geometric mean of 3 positive-cost ratios, one equal weight per benchmark. This benchmark set can include related tests and indexes; it is not a workload cost estimate.
DeepSeek-V4.1-Flash 7× cheaper
DeepSeek-V4.1-Flash 5.8× cheaper
DeepSeek-V4.1-Flash 3.1× cheaper
Settings or sources differ
DeepSeek-V4.1-Flash 7× cheaper
DeepSeek-V4.1-Flash 5.8× cheaper
DeepSeek-V4.1-Flash 3.1× cheaper
Settings or sources differ
Effort, protocol, aggregation and source match; other run conditions may differ.
DeepSeek-V4.1-Flash
Qwen3.8-27B
Effort, protocol, aggregation and source match; other run conditions may differ.
DeepSeek-V4.1-Flash
Qwen3.8-27B
Effort, protocol, aggregation and source match; other run conditions may differ.
DeepSeek-V4.1-Flash
Qwen3.8-27B
Settings or sources differ or are not recorded; interpret the comparison with care.
DeepSeek-V4.1-Flash
Qwen3.8-27B
| Benchmark | DeepSeek-V4.1-Flash | Qwen3.8-27B |
|---|---|---|
| AA Intelligence Index v5.0.1 | Not reported | 1 more settings |
| AA-LCR | Not reported | |
| Agent Arena | $0.1479 / task Median · Net improvement index (percentage points vs. average model): 2.3 Max (With Tools + Internet) | Not reported |
| CWE-bench v1 | Not reported | |
| FrontierCode 2.0 | $0.274 / task Average · Weighted rubric score: 18.2 Thinking High (With Tools + Internet) | Not reported |
| FrontierCS 2 | Not reported | |
| Terminal-Bench 4.0 | $0.4995 / task Average · Resolution rate (%): 17.2 Thinking High (With Tools) · vals-effort-study $0.4742 / task Average · Resolution rate (%): 16.2 Thinking Low (With Tools) · vals-effort-study 1 more settings$0.50 / task Average · Resolution rate (%): 19.7 Unspecified Reasoning (With Tools) · vals-mini-swe | Not reported |
| Vals Index v2.1 | Not reported | |
| Vibe Code Bench 1-100 | Not reported | |
| WeirdML v3 | $0.7371 / task Average · Official score (0-1; higher is better): 0.059 Thinking High (With Tools) | Not reported |
| Attribute | DeepSeek-V4.1-Flash | Qwen3.8-27B |
|---|---|---|
| Availability | Retired | Available |
| Provider | DeepSeek-AI | Alibaba |
| Model type | Multimodal model | Reasoning model |
| First release | September 10, 2026 | August 14, 2026 |
| Context window | 1M tokens | 256K tokens |
| Max output | 384K tokens | 128K tokens |
| Total parameters | 552B | 27B |
| Active parameters | 16B | 27B |
| Architecture | Mixture of Experts (MoE) | Dense |
| Documented capabilities | Reasoning, Multilingual, Coding, Function calling, Tool use, Structured output, Long context | Reasoning, Multilingual, Coding, Tool use, Function calling, Long context, Computer use |
| Input modalities | Text, Image | Text, Image, Video |
| Output modalities | Text | Text |
| Weights | Open weights | Open weights |
| Licence | MIT License | Apache 2.0 |
| Commercial use | Commercial use allowed | Commercial use allowed |
These are reported-score comparisons, not controlled head-to-head tests.
What people found comparing these models, or using either one. Comments also appear on the models' own pages.
Used DeepSeek-V4.1-Flash or Qwen3.8-27B? Sign in with Google to share what worked, what did not, and which API or setup you tested.
No community comments on either model yet.
Alternatives with the most benchmark overlap for this matchup.