Across 13 comparable benchmarks, DeepSeek-V4.1-Flash has better reported results on 10 and worse on 3. 1 further shared result is listed below, each with the reason it is not counted.
Moderate evidence13 of 14 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: Standard rates vs listed rates. The price section below shows the applicable periods.
| Fact | GPT-5.1 | DeepSeek-V4.1-Flash |
|---|---|---|
| Context window | 400K tokens | 1M tokens |
| Input / 1M | $1.25 | β |
| Output / 1M | $10.00 | β |
| Blended (4:1) | $2.19 | β |
| Status | Deprecated | Retired |
| Released | November 12, 2025 | September 10, 2026 |
Blended is 4:1 input to output with 90% cache hits at listed rates β set your own mix.
Mean of 8 shared benchmarks grouped by capability on a 0β100 scale. How averages are built
DeepSeek-V4.1-Flash ahead on 1 of 1 counted
DeepSeek-V4.1-Flash ahead on 2 of 2 counted
DeepSeek-V4.1-Flash ahead on 1 of 1 counted
Split 2β2 of 4 counted
GPT-5.1 ahead on 1 of 1 counted
DeepSeek-V4.1-Flash ahead on 1 of 1 counted
No shared comparable results for Computer & tool use, Research & search, Long context, Cybersecurity.
Published scores, with matching modes where available. Gaps show GPT-5.1 relative to DeepSeek-V4.1-Flash. Reading the results
| Benchmark | GPT-5.1 | DeepSeek-V4.1-Flash | Gap |
|---|---|---|---|
| Vibe Code Bench v1.1Code Generation & Editing Β· Pass rate (%) | 24.6Thinking High Β· tools | 84.7Thinking High Β· tools | β60.1 pts#82 vs #23 of 121 |
| ARC-AGI-1Abstract Generalization Β· Score (%) | 72.8Thinking High Β· no tools | 94.5β Max Β· no tools | β21.7 pts#51 vs #21 of 87 |
| ARC-AGI-2Abstract Generalization Β· Score (% solved); cost per task (USD)Not counted Β· Speed and cost telemetry is reported, not scored | 17.6Thinking High Β· no tools | 72.9β Max Β· no tools | 0.24Γ#55 vs #26 of 80 |
| SimpleBenchCommonsense Β· Score (AVG@5) | 53.2Thinking High Β· no tools | 66.7β Historical | 0.80Γ#59 vs #33 of 106 |
| MathArena ApexMathematics Β· Score (%) | 1.0Thinking High Β· tools | 65.6β Max Β· no tools | β64.6 pts#13 vs #4 of 16 |
| CritPtScientific Reasoning Β· Score | 4.9Thinking High Β· no tools | 14.3β Max Β· no tools | 0.34Γ#71 vs #43 of 134 |
| GPQA DiamondScientific Reasoning Β· Accuracy | 88.1Thinking Β· no tools | 90.9β Max Β· no tools | β2.8 pts#65 vs #44 of 200 |
| MedCodeClinical Workflows Β· Accuracy (%) | 52.7Thinking High Β· tools | 41.2Thinking High Β· tools | +11.6 pts#13 vs #54 of 97 |
| MedScribeClinical Workflows Β· Rubric pass rate (%) | 88.1Thinking High Β· tools | 85.5Thinking High Β· tools | +2.6 pts#14 vs #26 of 99 |
| MMMU-ProVisual Understanding Β· Accuracy | 79.0Thinking Β· no tools | 77.0β Max Β· no tools | +2.0 pts#38 vs #52 of 136 |
| SAGEService Workflows Β· Accuracy (%) | 43.2Thinking High Β· tools | 47.9Thinking High Β· tools | β4.7 pts#51 vs #29 of 82 |
| HLEKnowledge Exams Β· Accuracy | 42.7Thinking High Β· tools + web | 63.9β Max Β· tools | β21.2 pts#74 vs #7 of 151 |
| ECICapability Indices Β· ECI score (capability index, higher is better) | 150Reported best | 155Reported best | 0.97Γ#51 vs #30 of 168 |
| LiveBenchCross-capability Suites Β· Accuracy | 72.0Thinking High Β· no tools | 81.1β Max Β· no tools | β9.1 pts#50 vs #8 of 106 |
Code Generation & Editing Β· Pass rate (%)
Abstract Generalization Β· Score (%)
Abstract Generalization Β· Score (% solved); cost per task (USD)
Not counted Β· Speed and cost telemetry is reported, not scored
Commonsense Β· Score (AVG@5)
Mathematics Β· Score (%)
Scientific Reasoning Β· Score
Scientific Reasoning Β· Accuracy
Clinical Workflows Β· Accuracy (%)
Clinical Workflows Β· Rubric pass rate (%)
Visual Understanding Β· Accuracy
Service Workflows Β· Accuracy (%)
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 Β· 5% steps
Blended price = effective input rate at 90% cache hits Γ 80% + output rate Γ 20%
A blended cost needs a published input and output price on both sides, and one of these two does not have them yet.
GPT-5.1: Standard pricing. OpenAI official pricing (developers.openai.com/api/docs/pricing) verified 2026-11-13; standard tier; Deprecated; shutdown 2027-04-01 (replacement: gpt-6-sol)
DeepSeek-V4.1-Flash: Listed text rates; pricing tier not confirmed.
| Price / 1M tokens | GPT-5.1 | DeepSeek-V4.1-Flash | Ratio |
|---|---|---|---|
| Input | $1.25 | β | β |
| Output | $10.00 | β | β |
| Cached input | $0.125 | β | β |
| Cache write | β | β | β |
| Blended | $2.19 | β | β |
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.
GPT-5.1 relative to DeepSeek-V4.1-Flash
Summary: 2 of 2 shared benchmarks have matching recorded settings and sources. The average is a geometric mean of 2 positive-cost ratios, one equal weight per benchmark. This benchmark set can include related tests and indexes; it is not a workload cost estimate.
Effort, protocol, aggregation and source match; other run conditions may differ.
GPT-5.1
DeepSeek-V4.1-Flash
Effort, protocol, aggregation and source match; other run conditions may differ.
GPT-5.1
DeepSeek-V4.1-Flash
| Benchmark | GPT-5.1 | DeepSeek-V4.1-Flash |
|---|---|---|
| AA-LCR | Not reported | |
| Agent Arena | Not reported | $0.1479 / task Median Β· Net improvement index (percentage points vs. average model): 2.3 Max (With Tools + Internet) |
| CWE-bench v1 | Not reported | |
| FrontierCode 2.0 | Not reported | $0.274 / task Average Β· Weighted rubric score: 18.2 Thinking High (With Tools + Internet) $0.206 / task Average Β· Weighted rubric score: 16.0 Thinking Low (With Tools + Internet) |
| FrontierCS 2 | Not reported | |
| LiveBench 2.0 | Not reported | $0.1669 / task Average Β· Overall score (equal-weight mean of 7 suite scores, 0-100): 78.0 Max (task-specific tools) |
| MedCode | Not reported | |
| MedScribe | Not reported | |
| SAGE | Not reported | |
| Tax Agent Bench | Not reported | $0.1347 / task Average Β· Accuracy (%, rubric Γ citation quality): 26.6 Max (With Tools) |
| Terminal-Bench 4.0 | Not reported | $0.7666 / task Average Β· Resolution rate (%): 35.4 Max (With Tools) Β· vals-effort-study $0.4995 / task Average Β· Resolution rate (%): 17.2 Thinking High (With Tools) Β· vals-effort-study $0.4742 / task Average Β· Resolution rate (%): 16.2 Thinking Low (With Tools) Β· vals-effort-study 1 more settings$0.50 / task Average Β· Resolution rate (%): 19.7 Unspecified Reasoning (With Tools) Β· vals-mini-swe |
| Vals Index v2.1 | Not reported | |
| Vibe Code Bench 1-100 | Not reported | |
| Vibe Code Bench v1.1 | Not reported | |
| WeirdML v3 | Not reported | $0.7371 / task Average Β· Official score (0-1; higher is better): 0.059 Thinking High (With Tools) |
| Attribute | GPT-5.1 | DeepSeek-V4.1-Flash |
|---|---|---|
| Availability | Deprecated | Retired |
| Provider | OpenAI | DeepSeek-AI |
| Model type | Reasoning model | Multimodal model |
| First release | November 12, 2025 | September 10, 2026 |
| Context window | 400K tokens | 1M tokens |
| Max output | 128K tokens | 384K tokens |
| Total parameters | β | 552B |
| Active parameters | β | 16B |
| Architecture | Undisclosed | Mixture of Experts (MoE) |
| Documented capabilities | Reasoning, Multilingual | Reasoning, Multilingual, Coding, Function calling, Tool use, Structured output, Long context |
| Input modalities | Text, Image | Text, Image |
| Output modalities | Text | Text |
| Weights | Closed / API only | Open weights |
| Licence | β | MIT License |
| Commercial use | Proprietary | 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 GPT-5.1 or DeepSeek-V4.1-Flash? 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.