modelbench日本語

DeepSeek-V4.1-Flash

DeepSeekLegacy

Input price
—
Per 1 million tokens (USD)
Output price
—
Per 1 million tokens (USD)
Effort levels
not collected
Benchmark results
20total
Benchmark types: 20

Model information

API model ID
—
release date
2026-09-10
effort
not collected
Effort source
Not collected: the API effort list for this version has not been confirmed (API deepseek-flash returns the 1017 version from 2026-10-17). The HF card says reasoning effort is an integer from 1 to 100, and the table value is reasoning_effort=100.
Claude Code
unavailable
Codex
unavailable
Availability notes
Both are false because the model appears in neither the Claude Code official model list (code.claude.com/docs/en/model-config alias table, retrieved 2026-11-13; original: the table for "Use a model alias to select model settings without remembering exact version numbers" contains only Anthropic models such as fable, sonnet, opus, and haiku) nor the Codex official model list (learn.chatgpt.com/docs/models, retrieved 2026-11-13; original: "Recommended models" contains only the GPT-6 series) (connections via a custom provider are not counted as official options, and connections via a gateway are not counted as official options).
Availability source
https://code.claude.com/docs/en/model-config

Price (USD per 1 million tokens)

Prices do not vary by effort; token consumption does.

input—
output—
cache read—
cache write—

Pricing for this version alone is not collected (the price at the 2026-09-10 announcement may have carried over to the 1017 version, but this has not been confirmed).

Benchmark results

BenchmarkeffortScoreSource cost (metric)measured/reported bysourceConditions and extraction method
Terminal-Bench 4.0not statednot stated31.2%—DeepSeekDeepSeek-V4.1-Flash (Hugging Face model card, deepseek-ai)official · Retrieved 2026-11-13Vendor-reported value (self-reported by DeepSeek). Card "Instruct Model": All instruct results below use the maximum effort setting (reasoning_effort=100), temperature=1.0, top_p=0.95. The correspondence of reasoning_effort=100 to the API effort names (low/high/max) has no official statement, so effort is null. Evaluated in Minimal mode of DeepSeek Harness with 1M context (the DeepSeek-V4.1-Flash-1017 card lists the same 31.2 in the Claude Code harness table, so the harness descriptions differ).transcribed from table
Source excerptTerminal-Bench 4.0 (Pass@1) | 31.2Location HF model card "Comparison with frontier models (Max reasoning effort)" table, Terminal-Bench 4.0 (Pass@1) row, DS-V4.1-Flash column (rightmost)
Terminal-Bench 3.0not statednot stated30.0%—DeepSeekDeepSeek-V4.1-Flash (Hugging Face model card, deepseek-ai)official · Retrieved 2026-11-13Vendor-reported value (self-reported by DeepSeek). Card "Instruct Model": All instruct results below use the maximum effort setting (reasoning_effort=100), temperature=1.0, top_p=0.95. The correspondence of reasoning_effort=100 to the API effort names (low/high/max) has no official statement, so effort is null. Evaluated in Minimal mode of DeepSeek Harness with 1M context.transcribed from table
Source excerptTerminal-Bench 3.0 (Pass@1) | 30.0Location HF model card "Comparison with frontier models (Max reasoning effort)" table, Terminal-Bench 3.0 (Pass@1) row, DS-V4.1-Flash column (rightmost)
Terminal-Bench 2.1not statednot stated90.6%—DeepSeekDeepSeek-V4.1-Flash (Hugging Face model card, deepseek-ai)official · Retrieved 2026-11-13Vendor-reported value (self-reported by DeepSeek). Card "Instruct Model": All instruct results below use the maximum effort setting (reasoning_effort=100), temperature=1.0, top_p=0.95. The correspondence of reasoning_effort=100 to the API effort names (low/high/max) has no official statement, so effort is null. Evaluated in Minimal mode of DeepSeek Harness with 1M context.transcribed from table
Source excerptTerminal-Bench 2.1 (Pass@1) | 90.6Location HF model card "Comparison with frontier models (Max reasoning effort)" table, Terminal-Bench 2.1 (Pass@1) row, DS-V4.1-Flash column (rightmost)
DeepSWE v1.1not statednot stated74.2%—DeepSeekDeepSeek-V4.1-Flash (Hugging Face model card, deepseek-ai)official · Retrieved 2026-11-13Vendor-reported value (self-reported by DeepSeek). Card "Instruct Model": All instruct results below use the maximum effort setting (reasoning_effort=100), temperature=1.0, top_p=0.95. The correspondence of reasoning_effort=100 to the API effort names (low/high/max) has no official statement, so effort is null. Evaluated with the mini-SWE harness.transcribed from table
Source excerptDeepSWE v1.1 (Resolved) | 74.2Location HF model card "Comparison with frontier models (Max reasoning effort)" table, DeepSWE v1.1 (Resolved) row, DS-V4.1-Flash column (rightmost)
Agents' Last Exam (pass rate, version not stated)not statednot stated31.8%—DeepSeekDeepSeek-V4.1-Flash (Hugging Face model card, deepseek-ai)official · Retrieved 2026-11-13Vendor-reported value (self-reported by DeepSeek). Card "Instruct Model": All instruct results below use the maximum effort setting (reasoning_effort=100), temperature=1.0, top_p=0.95. The correspondence of reasoning_effort=100 to the API effort names (low/high/max) is not stated officially, so effort is null. Evaluated using the official scaffold. Version not stated.transcribed from table
Source excerptAgent's Last Exam (Pass@1) | 31.8Location HF model card "Comparison with frontier models (Max reasoning effort)" table, Agent's Last Exam (Pass@1) row, DS-V4.1-Flash column (rightmost)
Chartography (with tools)not statednot stated78.9%—DeepSeekDeepSeek-V4.1-Flash (Hugging Face model card, deepseek-ai)official · Retrieved 2026-11-13Vendor-reported value (self-reported by DeepSeek). Card "Instruct Model": All instruct results below use the maximum effort setting (reasoning_effort=100), temperature=1.0, top_p=0.95. The correspondence of reasoning_effort=100 to the API effort names (low/high/max) has no official statement, so effort is null. Evaluated with the Claude Code harness and 512k context.transcribed from table
Source excerptChartography w/ tools (Pass@1) | 78.9Location HF model card "Comparison with frontier models (Max reasoning effort)" table, Chartography w/ tools (Pass@1) row, DS-V4.1-Flash column (rightmost)
SEC-bench Pronot statednot stated62.8%—DeepSeekDeepSeek-V4.1-Flash (Hugging Face model card, deepseek-ai)official · Retrieved 2026-11-13Vendor-reported value (self-reported by DeepSeek). Card "Instruct Model": All instruct results below use the maximum effort setting (reasoning_effort=100), temperature=1.0, top_p=0.95. The correspondence of reasoning_effort=100 to the API effort names (low/high/max) has no official statement, so effort is null. Evaluated with the Claude Code harness.transcribed from table
Source excerptSEC-Bench Pro (Pass@1) | 62.8Location HF model card "Comparison with frontier models (Max reasoning effort)" table, SEC-Bench Pro (Pass@1) row, DS-V4.1-Flash column (rightmost)
GPQA Diamondnot statednot stated89.8%—Epoch AIEpoch AI, Capabilities & benchmarking (benchmark_data.zip, gpqa_diamond.csv) CC BY 4.0third-party · Retrieved 2026-11-12Run by Epoch AI (third-party measurement). Source label "DeepSeek V4.1 Flash (max)". effort is indicated by the suffix _max in Epoch's Model version. Data retrieved 2026-11-12. No effort for the source effort label "max" because the official effort list has not been collected.transcribed from table
Source excerptgpqa_diamond.csv: Model version=deepseek-v4.1-flash_max, Best score (across scorers)=0.8983585858585859, stderr=0.017287759372442393, Started at=2026-10-09T14:38:01.000ZLocation gpqa_diamond.csv, data row 16 (file lines 39-40; multi-line record), id=ZU5ebpDaDPiDZLgv7cG8B9, Model version=deepseek-v4.1-flash_max
FrontierMath Tiers 1-3(v2)not statednot stated67.4%—Epoch AIEpoch AI, Capabilities & benchmarking (benchmark_data.zip, frontiermath_tiers_1_3_v2.csv) CC BY 4.0third-party · Retrieved 2026-11-12Run by Epoch AI (third-party measurement). Source label "DeepSeek V4.1 Flash (max)". effort is indicated by the suffix _max in Epoch's Model version. Data retrieved 2026-11-12. No effort for the source effort label "max" because the official effort list has not been collected.transcribed from table
Source excerptfrontiermath_tiers_1_3_v2.csv: Model version=deepseek-v4.1-flash_max, Best score (across scorers)=0.6736842105263158, stderr=0.027821981570324446, Started at=2026-10-08T19:02:32.000ZLocation frontiermath_tiers_1_3_v2.csv, data row 10 (file lines 22-23; multi-line record), id=b6dXtaGSwvfRYADKYnj9Ci, Model version=deepseek-v4.1-flash_max
FrontierMath Tier 4(v2)not statednot stated26.8%—Epoch AIEpoch AI, Capabilities & benchmarking (benchmark_data.zip, frontiermath_tier_4_v2.csv) CC BY 4.0third-party · Retrieved 2026-11-12Run by Epoch AI (third-party measurement). Source label "DeepSeek V4.1 Flash (max)". effort is indicated by the suffix _max in Epoch's Model version. Data retrieved 2026-11-12. No effort for the source effort label "max" because the official effort list has not been collected.transcribed from table
Source excerptfrontiermath_tier_4_v2.csv: Model version=deepseek-v4.1-flash_max, Best score (across scorers)=0.2682926829268293, stderr=0.07005564203095158, Started at=2026-10-08T19:02:32.000ZLocation frontiermath_tier_4_v2.csv, data row 10 (file lines 22-23; multi-line record), id=KansJcyeGR4m7G4p8GupDn, Model version=deepseek-v4.1-flash_max
OTIS Mock AIME 2024-2025not statednot stated98.3%—Epoch AIEpoch AI, Capabilities & benchmarking (benchmark_data.zip, otis_mock_aime_2024_2025.csv) CC BY 4.0third-party · Retrieved 2026-11-12Run by Epoch AI (third-party measurement). Source label "DeepSeek V4.1 Flash (max)". effort is indicated by the suffix _max in Epoch's Model version. Data retrieved 2026-11-12. No effort for the source effort label "max" because the official effort list has not been collected.transcribed from table
Source excerptotis_mock_aime_2024_2025.csv: Model version=deepseek-v4.1-flash_max, Best score (across scorers)=0.9833333333333333, stderr=0.006405884289881728, Started at=2026-10-08T19:02:32.000ZLocation otis_mock_aime_2024_2025.csv, data row 17 (file lines 40-41; multi-line record), id=7eHNiVQCzLcAyj5LxmeeSU, Model version=deepseek-v4.1-flash_max
SimpleQA Verifiednot statednot stated42.9%—Epoch AIEpoch AI, Capabilities & benchmarking (benchmark_data.zip, simpleqa_verified.csv) CC BY 4.0third-party · Retrieved 2026-11-12Run by Epoch AI (third-party measurement). Source label "DeepSeek V4.1 Flash (max)". effort is indicated by the suffix _max in Epoch's Model version. Data retrieved 2026-11-12. No effort for the source effort label "max" because the official effort list has not been collected.transcribed from table
Source excerptsimpleqa_verified.csv: Model version=deepseek-v4.1-flash_max, Best score (across scorers)=0.429, stderr=0.015658997547870177, Started at=2026-10-13T12:59:33.000ZLocation simpleqa_verified.csv, data row 15 (file lines 27-28; multi-line record), id=3JJ4ChNxMo8CGAh3HrKNrw, Model version=deepseek-v4.1-flash_max
Chess Puzzles(Epoch AI)not statednot stated48.0%—Epoch AIEpoch AI, Capabilities & benchmarking (benchmark_data.zip, chess_puzzles.csv) CC BY 4.0third-party · Retrieved 2026-11-12Run by Epoch AI (third-party measurement). Source label "DeepSeek V4.1 Flash (max)". effort is indicated by the suffix _max in Epoch's Model version. Data retrieved 2026-11-12. No effort for the source effort label "max" because the official effort list has not been collected.transcribed from table
Source excerptchess_puzzles.csv: Model version=deepseek-v4.1-flash_max, Best score (across scorers)=0.48, stderr=0.05021167315686782, Started at=2026-10-13T12:39:44.000ZLocation chess_puzzles.csv, data row 160 (file lines 216-217; multi-line record), id=ACpDTbhRPwce2BWQTWD8iC, Model version=deepseek-v4.1-flash_max
MirrorCodenot statednot stated64.4%—Epoch AIEpoch AI, Capabilities & benchmarking (benchmark_data.zip, mirrorcode.csv) CC BY 4.0third-party · Retrieved 2026-11-12Run by Epoch AI (third-party measurement). Source label "DeepSeek V4.1 Flash (max)". effort is indicated by the suffix _max in Epoch's Model version. Data retrieved 2026-11-12. No effort for the source effort label "max" because the official effort list has not been collected.transcribed from table
Source excerptmirrorcode.csv: Model version=deepseek-v4.1-flash_max, Best score (across scorers)=0.6444444444444445, stderr=0.11369999821616648, Started at=2026-10-21T16:49:36.000ZLocation mirrorcode.csv, data row 11 (file lines 15-16; multi-line record), id=nRrYnwt7v5xuS4K29sTubp, Model version=deepseek-v4.1-flash_max
Epoch Capabilities Index(ECI)not statednot stated156.7—Epoch AIEpoch AI, epoch_capabilities_index/eci_scores.csv (CC BY 4.0)third-party · Retrieved 2026-11-12A per-model metric that combines results across multiple benchmarks and multiple efforts (not per effort). Calculated by Epoch AI.transcribed from table
Source excerpteci_scores.csv: Model=DeepSeek V4.1 Flash, eci=156.7, eci_ci_low=154.79, eci_ci_high=158.78Location epoch_capabilities_index/eci_scores.csv, data row 34 (file line 35), Model=DeepSeek V4.1 Flash
ProgramBench(Almost Solved)not statednot stated20.3%—DeepSeekDeepSeek-V4.1-Flash (Hugging Face model card, deepseek-ai)official · Retrieved 2026-11-13Vendor-reported value (self-reported by DeepSeek). Card "Instruct Model": All instruct results below use the maximum effort setting (reasoning_effort=100), temperature=1.0, top_p=0.95. The correspondence of reasoning_effort=100 to the API effort names (low/high/max) has no official statement, so effort is null. Evaluated in Minimal mode of DeepSeek Harness.transcribed from table
Source excerptProgramBench (Almost@1) | 20.3Location HF model card "Comparison with frontier models (Max reasoning effort)" table, ProgramBench (Almost@1) row, DS-V4.1-Flash column (rightmost)
NL2Repo-Benchnot statednot stated64—DeepSeekDeepSeek-V4.1-Flash (Hugging Face model card, deepseek-ai)official · Retrieved 2026-11-13Vendor-reported value (self-reported by DeepSeek). Card "Instruct Model": All instruct results below use the maximum effort setting (reasoning_effort=100), temperature=1.0, top_p=0.95. The correspondence of reasoning_effort=100 to the API effort names (low/high/max) has no official statement, so effort is null. Evaluated in Minimal mode of DeepSeek Harness. The comparison table in the HF model card for DeepSeek-V4.1-Flash-1017 gives 65.4 for NL2Repo-Bench on the same DS-V4.1-Flash, differing from 64.0 in this card (the value follows this card).transcribed from table
Source excerptNL2Repo-Bench (Score) | 64.0Location HF model card "Comparison with frontier models (Max reasoning effort)" table, NL2Repo-Bench (Score) row, DS-V4.1-Flash column (rightmost)
AutomationBench v1.0.6not statednot stated54.8%—DeepSeekDeepSeek-V4.1-Flash (Hugging Face model card, deepseek-ai)official · Retrieved 2026-11-13Vendor-reported value (self-reported by DeepSeek). Card "Instruct Model": All instruct results below use the maximum effort setting (reasoning_effort=100), temperature=1.0, top_p=0.95. The correspondence of reasoning_effort=100 to the API effort names (low/high/max) is not stated officially, so effort is null. Evaluated using the official scaffold. The version is absent from this card, but the table in the DeepSeek-V4.1-Flash-1017 card lists the same 54.8 in the "Automation-Bench v1.0.6" row.transcribed from table
Source excerptAutomationBench (Pass@1) | 54.8Location HF model card "Comparison with frontier models (Max reasoning effort)" table, AutomationBench (Pass@1) row, DS-V4.1-Flash column (rightmost)
BabyVision (with tools)not statednot stated89.6%—DeepSeekDeepSeek-V4.1-Flash (Hugging Face model card, deepseek-ai)official · Retrieved 2026-11-13Vendor-reported value (self-reported by DeepSeek). Card "Instruct Model": All instruct results below use the maximum effort setting (reasoning_effort=100), temperature=1.0, top_p=0.95. The correspondence of reasoning_effort=100 to the API effort names (low/high/max) has no official statement, so effort is null. Evaluated with the Claude Code harness and 512k context.transcribed from table
Source excerptBabyVision w/ tools (Pass@1) | 89.6Location HF model card "Comparison with frontier models (Max reasoning effort)" table, BabyVision w/ tools (Pass@1) row, DS-V4.1-Flash column (rightmost)
ZeroBench main (with tools, pass@5)not statednot stated49.0%—DeepSeekDeepSeek-V4.1-Flash (Hugging Face model card, deepseek-ai)official · Retrieved 2026-11-13Vendor-reported value (self-reported by DeepSeek). Card "Instruct Model": All instruct results below use the maximum effort setting (reasoning_effort=100), temperature=1.0, top_p=0.95. The correspondence of reasoning_effort=100 to the API effort names (low/high/max) has no official statement, so effort is null. Evaluated with the Claude Code harness and 512k context.transcribed from table
Source excerptZeroBench-main w/ tools (Pass@5) | 49.0Location HF model card "Comparison with frontier models (Max reasoning effort)" table, ZeroBench-main w/ tools (Pass@5) row, DS-V4.1-Flash column (rightmost)

Related rankings

References