# DeepSeek V4 FlashMax > Source: Benchize (https://benchize.com) — https://benchize.com/models/deepseek-v4-flash-max. When citing or reusing this content, credit "Benchize" and link to https://benchize.com/models/deepseek-v4-flash-max. DeepSeek Canonical URL: https://benchize.com/models/deepseek-v4-flash-max Last updated: Thu, 22 Oct 2026 07:10:12 GMT Reproduction and quotation must name Benchize, link to the canonical HTML URL, and retain individual source citations. DeepSeek ## Benchmarks | Benchmark | Score | ± | Rank | Cost | Latency | Source | | | --- | --- | --- | --- | --- | --- | --- | --- | | [ArXivMath 05/2026](https://benchize.com/benchmarks/matharena%3Aarxiv--may.md) | 55.83% | 95% CI 46.94–64.72% | 16 / 23 | 0.04usd per problem | — | [matharena · arxiv--may](https://matharena.ai/competition_tables/arxiv--may) | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 55.83 \| percent \| competition \| Higher is better \| \| `cost_per_problem` \| 0.039 \| usd_per_problem \| competition \| Lower is better \| ### Configuration competition arxiv--may predicted false Reasoning effort max sourceModelName DeepSeek-v4-Flash (Max) sourceModelVersion deepseek/deepseek_v4_flash thinkingEnabled true ### Reported statistics Uncertainty 95% confidence interval: 46.94–64.72% confidenceInterval95 8.89 problems 40 ### Source [Original source](https://matharena.ai/competition_tables/arxiv--may) Source matharena Original model label DeepSeek-v4-Flash (Max) Source version arxiv--may Observed 2026-10-22T05:30:08.858Z Evidence SHA-256 `758e1bba77e36f4ced2a51b4bc0d46bd956ed94ef6f57e93e1a974b75e76a4ce` Source locator `/table/rows/16` | | | | | | | | | [BrokenArXiv 02/2026](https://benchize.com/benchmarks/matharena%3Aarxiv_false--february.md) | 17.74% | 95% CI 11.02–24.46% | 6 / 15 | 0.04usd per problem | — | [matharena · arxiv_false--february](https://matharena.ai/competition_tables/arxiv_false--february) | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 17.74 \| percent \| competition \| Higher is better \| \| `cost_per_problem` \| 0.036 \| usd_per_problem \| competition \| Lower is better \| ### Configuration competition arxiv_false--february predicted false Reasoning effort max sourceModelName DeepSeek-v4-Flash (Max) sourceModelVersion deepseek/deepseek_v4_flash thinkingEnabled true ### Reported statistics Uncertainty 95% confidence interval: 11.02–24.46% confidenceInterval95 6.72 problems 31 ### Source [Original source](https://matharena.ai/competition_tables/arxiv_false--february) Source matharena Original model label DeepSeek-v4-Flash (Max) Source version arxiv_false--february Observed 2026-10-22T05:30:05.567Z Evidence SHA-256 `a6bc6fc18da7dd181ed727da924f9569d3e323c31c1aba9a4d9fa3bff492bee3` Source locator `/table/rows/5` | | | | | | | | | [ArXivMath 02/2026](https://benchize.com/benchmarks/matharena%3Aarxiv--february.md) | 43.75% | 95% CI 35.16–52.34% | 9 / 20 | 0.05usd per problem | — | [matharena · arxiv--february](https://matharena.ai/competition_tables/arxiv--february) | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 43.75 \| percent \| competition \| Higher is better \| \| `cost_per_problem` \| 0.049 \| usd_per_problem \| competition \| Lower is better \| ### Configuration competition arxiv--february predicted false Reasoning effort max sourceModelName DeepSeek-v4-Flash (Max) sourceModelVersion deepseek/deepseek_v4_flash thinkingEnabled true ### Reported statistics Uncertainty 95% confidence interval: 35.16–52.34% confidenceInterval95 8.59 problems 32 ### Source [Original source](https://matharena.ai/competition_tables/arxiv--february) Source matharena Original model label DeepSeek-v4-Flash (Max) Source version arxiv--february Observed 2026-10-22T05:30:08.096Z Evidence SHA-256 `3bc5b0a1eb91a0bdaa3d0942ebf82327a8664334b225ae3b2dc1d8b4f57efa51` Source locator `/table/rows/8` | | | | | | | | | [AIME 2026](https://benchize.com/benchmarks/aime-2026.md) | 95.83% | 95% CI 92.25–99.41% | 12 / 21 | 0.009usd per problem | — | [matharena · aime--aime_2026](https://matharena.ai/competition_tables/aime--aime_2026) | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 95.83 \| percent \| competition \| Higher is better \| \| `cost_per_problem` \| 0.009 \| usd_per_problem \| competition \| Lower is better \| ### Configuration competition aime--aime_2026 predicted false Reasoning effort max sourceModelName DeepSeek-v4-Flash (Max) sourceModelVersion deepseek/deepseek_v4_flash thinkingEnabled true ### Reported statistics Uncertainty 95% confidence interval: 92.25–99.41% confidenceInterval95 3.58 problems 30 ### Source [Original source](https://matharena.ai/competition_tables/aime--aime_2026) Source matharena Original model label DeepSeek-v4-Flash (Max) Source version aime--aime_2026 Observed 2026-10-22T05:30:13.366Z Evidence SHA-256 `58b3ac1e0805975eba566c493b8e0165d41eb7ca517218106e112529fb6342de` Source locator `/table/rows/12` | | | | | | | | | [TapTap Maker heldout](https://benchize.com/benchmarks/taptap-maker%3Aheldout.md) | 47.83% | Range 47.83–47.83% | 26 / 30 | — | — | [taptap maker](https://maker.taptap.cn/leaderboard/data/editions/2026-10-18-v6-voxel.json) | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `l2_pass_rate` \| 0.4783 \| score_fraction \| reported \| Higher is better \| ### Configuration datasetVersion 0.2.3 edition 2026-10-18-v6-voxel harness claude Reasoning effort max reasoningMode Not reported ### Reported statistics Uncertainty Reported interval: 47.83–47.83% interval [0.4783,0.4783] intervalMethod bootstrap-case rank 30 rankInterval [29,30] ### Source [Original source](https://maker.taptap.cn/leaderboard/data/editions/2026-10-18-v6-voxel.json) Source taptap maker Original model label deepseek-v4-flash Observed 2026-10-22T06:30:04.044Z Evidence SHA-256 `ad2d7f1adeae0548a4edc0b5e75d11449b443449ff76533d04bc1fb1b6d7035e` Source locator `/heldout/29` | | | | | | | | | [ArXivMath 03/2026](https://benchize.com/benchmarks/matharena%3Aarxiv--march.md) | 52.5% | 95% CI 43.57–61.43% | 9 / 14 | 0.04usd per problem | — | [matharena · arxiv--march](https://matharena.ai/competition_tables/arxiv--march) | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 52.50 \| percent \| competition \| Higher is better \| \| `cost_per_problem` \| 0.038 \| usd_per_problem \| competition \| Lower is better \| ### Configuration competition arxiv--march predicted false Reasoning effort max sourceModelName DeepSeek-v4-Flash (Max) sourceModelVersion deepseek/deepseek_v4_flash thinkingEnabled true ### Reported statistics Uncertainty 95% confidence interval: 43.57–61.43% confidenceInterval95 8.93 problems 30 ### Source [Original source](https://matharena.ai/competition_tables/arxiv--march) Source matharena Original model label DeepSeek-v4-Flash (Max) Source version arxiv--march Observed 2026-10-22T05:30:08.323Z Evidence SHA-256 `9cb6d9cc26949fcb2b6affed792584b4bfe9d6168e1fc5cf2c13b38d3479715b` Source locator `/table/rows/9` | | | | | | | | | [Apex](https://benchize.com/benchmarks/matharena%3Aapex--apex_2025.md) | 27.08% | 95% CI 17.8–36.36% | 10 / 31 | 0.05usd per problem | — | [matharena · apex--apex_2025](https://matharena.ai/competition_tables/apex--apex_2025) | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 27.08 \| percent \| competition \| Higher is better \| \| `cost_per_problem` \| 0.050 \| usd_per_problem \| competition \| Lower is better \| ### Configuration competition apex--apex_2025 predicted false Reasoning effort max sourceModelName DeepSeek-v4-Flash (Max) sourceModelVersion deepseek/deepseek_v4_flash thinkingEnabled true ### Reported statistics Uncertainty 95% confidence interval: 17.8–36.36% confidenceInterval95 9.28 problems 12 ### Source [Original source](https://matharena.ai/competition_tables/apex--apex_2025) Source matharena Original model label DeepSeek-v4-Flash (Max) Source version apex--apex_2025 Observed 2026-10-22T05:30:13.849Z Evidence SHA-256 `20b28370d9d7f5a15a19fc9e4226c9857cdb07bc42b32ccec508e687b7b9f3fc` Source locator `/table/rows/10` | | | | | | | | | [HMMT Feb 2026](https://benchize.com/benchmarks/matharena%3Ahmmt--hmmt_feb_2026.md) | 93.94% | 95% CI 89.87–98.01% | 8 / 21 | 0.02usd per problem | — | [matharena · hmmt--hmmt_feb_2026](https://matharena.ai/competition_tables/hmmt--hmmt_feb_2026) | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 93.94 \| percent \| competition \| Higher is better \| \| `cost_per_problem` \| 0.015 \| usd_per_problem \| competition \| Lower is better \| ### Configuration competition hmmt--hmmt_feb_2026 predicted false Reasoning effort max sourceModelName DeepSeek-v4-Flash (Max) sourceModelVersion deepseek/deepseek_v4_flash thinkingEnabled true ### Reported statistics Uncertainty 95% confidence interval: 89.87–98.01% confidenceInterval95 4.07 problems 33 ### Source [Original source](https://matharena.ai/competition_tables/hmmt--hmmt_feb_2026) Source matharena Original model label DeepSeek-v4-Flash (Max) Source version hmmt--hmmt_feb_2026 Observed 2026-10-22T05:30:13.595Z Evidence SHA-256 `0dce922c33a54d29d4a1872a42d74077b491915c80d8433f4b379ea0b78bb979` Source locator `/table/rows/9` | | | | | | | | | [AutomationBench](https://benchize.com/benchmarks/zapier%3Aautomationbench.md) | 18.11% | — | 43 / 113 | $0.04 | — | [zapier](https://framerusercontent.com/sites/4WTSl4BNjd1q9QFEFibC6h/lOvLnhItj8S8-1vyjH_s33Jn-WmUXPmiJ5Diajl0Q38.CMutaHYl.mjs) | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `score` \| 18.11 \| percent \| reported \| Higher is better \| \| `cost` \| 0.04 \| USD \| per_task \| Lower is better \| ### Configuration benchmarkVersion 1.0.6 costFootnote false Fallback policy Not reported Reasoning effort max Model configuration DeepSeek V4 Flash (Max) thinkingEnabled true ### Reported statistics rank 44 ### Source [Original source](https://framerusercontent.com/sites/4WTSl4BNjd1q9QFEFibC6h/lOvLnhItj8S8-1vyjH_s33Jn-WmUXPmiJ5Diajl0Q38.CMutaHYl.mjs) Source zapier Original model label DeepSeek V4 Flash (Max) Observed 2026-10-21T16:34:56.783Z Evidence SHA-256 `43978e38f018789fa59c11de5328abe9e5905bfa9a5a4d601fa3aa4d9c153b0d` Source locator `/rows/43` | | | | | | | | | [TapTap Maker public](https://benchize.com/benchmarks/taptap-maker%3Apublic.md) | 69.08% | Range 69.08–69.08% | 27 / 30 | $0.1 | — | [taptap maker](https://maker.taptap.cn/leaderboard/data/editions/2026-10-18-v6-voxel.json) | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `l2_pass_rate` \| 0.6908 \| score_fraction \| reported \| Higher is better \| \| `cost_per_case` \| 0.1013 \| USD \| per_case \| Lower is better \| ### Configuration datasetVersion 0.2.3 edition 2026-10-18-v6-voxel harness claude Reasoning effort max reasoningMode model-managed ### Reported statistics Uncertainty Reported interval: 69.08–69.08% interval [0.6908,0.6908] intervalMethod bootstrap-case rank 31 rankInterval [31,31] ### Source [Original source](https://maker.taptap.cn/leaderboard/data/editions/2026-10-18-v6-voxel.json) Source taptap maker Original model label deepseek-v4-flash Observed 2026-10-22T06:30:04.044Z Evidence SHA-256 `ad2d7f1adeae0548a4edc0b5e75d11449b443449ff76533d04bc1fb1b6d7035e` Source locator `/public/30` | | | | | | | | | [APEX-Agents](https://benchize.com/benchmarks/mercor%3Aapex-agents%3A1.1.md) | 55.3% | — | 15 / 38 | — | — | [mercor · 1.1](https://www.mercor.com/apex/apex-agents-leaderboard/) | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `pass_1` \| 55.3 \| percent \| task_completion_pass_at_1 \| Higher is better \| \| `mean_score` \| 67.5 \| percent \| rubric_mean \| Higher is better \| ### Configuration datasetUrl https://huggingface.co/datasets/mercor/apex-agents-v1.1 datasetVersion 1.1 harness loop_truncated_tools_agent Reasoning effort max ### Reported statistics reportedErrors {"mean_score":4.6,"pass_1":5.6} sampleSize 829 ### Source [Original source](https://www.mercor.com/apex/apex-agents-leaderboard/) Source mercor Original model label DeepSeek-V4-Flash Source version 1.1 Observed 2026-10-22T03:20:05.238Z Evidence SHA-256 `9a5cc96c60cc0788561a54a64689121504d26058f7690c1106f2490706183370` Source locator `/__NEXT_DATA__/props/pageProps/leaderboardData/36` | | | | | | | | | [ArXivMath 04/2026](https://benchize.com/benchmarks/matharena%3Aarxiv--april.md) | 47.5% | 95% CI 36.56–58.44% | 15 / 20 | 0.04usd per problem | — | [matharena · arxiv--april](https://matharena.ai/competition_tables/arxiv--april) | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 47.50 \| percent \| competition \| Higher is better \| \| `cost_per_problem` \| 0.038 \| usd_per_problem \| competition \| Lower is better \| ### Configuration competition arxiv--april predicted false Reasoning effort max sourceModelName DeepSeek-v4-Flash (Max) sourceModelVersion deepseek/deepseek_v4_flash thinkingEnabled true ### Reported statistics Uncertainty 95% confidence interval: 36.56–58.44% confidenceInterval95 10.94 problems 40 ### Source [Original source](https://matharena.ai/competition_tables/arxiv--april) Source matharena Original model label DeepSeek-v4-Flash (Max) Source version arxiv--april Observed 2026-10-22T05:30:08.600Z Evidence SHA-256 `5139c835884bfab38f18d23a248814af4680927310bc5cd2a86ea2eab779e0a7` Source locator `/table/rows/16` | | | | | | | | | [Apex Shortlist](https://benchize.com/benchmarks/matharena%3Aapex--shortlist_2025.md) | 89.36% | 95% CI 84.95–93.77% | 4 / 25 | 0.04usd per problem | — | [matharena · apex--shortlist_2025](https://matharena.ai/competition_tables/apex--shortlist_2025) | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 89.36 \| percent \| competition \| Higher is better \| \| `cost_per_problem` \| 0.040 \| usd_per_problem \| competition \| Lower is better \| ### Configuration competition apex--shortlist_2025 predicted false Reasoning effort max sourceModelName DeepSeek-v4-Flash (Max) sourceModelVersion deepseek/deepseek_v4_flash thinkingEnabled true ### Reported statistics Uncertainty 95% confidence interval: 84.95–93.77% confidenceInterval95 4.41 problems 47 ### Source [Original source](https://matharena.ai/competition_tables/apex--shortlist_2025) Source matharena Original model label DeepSeek-v4-Flash (Max) Source version apex--shortlist_2025 Observed 2026-10-22T05:30:14.141Z Evidence SHA-256 `74513abecacb4d230d741f4d6df425ce878fb450220e88a1a6823524476c2989` Source locator `/table/rows/4` | | | | | | | | | [DeepSWE](https://benchize.com/benchmarks/deepswe%3Av1.1.md) | 53.32% | Range 49.75–56.89% | 46 / 70 | $0.1 | 1,439s | [deepswe](https://deepswe.datacurve.ai/artifacts/v1.1/leaderboard-live.json) | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `pass_at_1` \| 0.5331858407079646 \| score_fraction \| scored_rollout_attempts \| Higher is better \| \| `pass_at_4` \| 0.8053097345132744 \| score_fraction \| tasks_with_one_passing_rollout \| Higher is better \| \| `cost_per_attempt` \| 0.10023560370265487 \| USD \| scored_attempt \| Lower is better \| \| `duration` \| 1438.9961937522125 \| seconds \| scored_attempt \| Lower is better \| ### Configuration harness mini-swe-agent Reasoning effort max sourceConfiguration mini_swe_agent_deepseek_v4_flash_max ### Reported statistics Uncertainty Reported interval: 49.75–56.89% interval [0.4975163385576403,0.5688553428582889] intervalMethod 95% run-to-run: SE across repeated whole-benchmark passes (1.96 * std(runs)/sqrt(R)) runCount 4 sampleSize 452 taskCount 113 ### Source [Original source](https://deepswe.datacurve.ai/artifacts/v1.1/leaderboard-live.json) Source deepswe Original model label deepseek-v4-flash Observed 2026-10-22T03:50:05.889Z Evidence SHA-256 `a7c15d66288fd249c020b9931c017b92d1a3b90e480b3ff34974b752bd030019` Source locator `/rows/45` | | | | | | | | | [BrokenArXiv 03/2026](https://benchize.com/benchmarks/matharena%3Aarxiv_false--march.md) | 19.64% | 95% CI 12.28–27% | 4 / 14 | 0.03usd per problem | — | [matharena · arxiv_false--march](https://matharena.ai/competition_tables/arxiv_false--march) | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 19.64 \| percent \| competition \| Higher is better \| \| `cost_per_problem` \| 0.031 \| usd_per_problem \| competition \| Lower is better \| ### Configuration competition arxiv_false--march predicted false Reasoning effort max sourceModelName DeepSeek-v4-Flash (Max) sourceModelVersion deepseek/deepseek_v4_flash thinkingEnabled true ### Reported statistics Uncertainty 95% confidence interval: 12.28–27% confidenceInterval95 7.36 problems 56 ### Source [Original source](https://matharena.ai/competition_tables/arxiv_false--march) Source matharena Original model label DeepSeek-v4-Flash (Max) Source version arxiv_false--march Observed 2026-10-22T05:30:05.839Z Evidence SHA-256 `b2ea8c773acad8b54736938938d6ee37fa533bfc5b985eefc303733bfcb0c6f0` Source locator `/table/rows/3` | | | | | | | | | [ArXivMath 06/2026](https://benchize.com/benchmarks/matharena%3Aarxiv--june.md) | 43.75% | 95% CI 35.65–51.85% | 25 / 27 | 0.04usd per problem | — | [matharena · arxiv--june](https://matharena.ai/competition_tables/arxiv--june) | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 43.75 \| percent \| competition \| Higher is better \| \| `cost_per_problem` \| 0.038 \| usd_per_problem \| competition \| Lower is better \| ### Configuration competition arxiv--june predicted false Reasoning effort max sourceModelName DeepSeek-v4-Flash (Max) sourceModelVersion deepseek/deepseek_v4_flash thinkingEnabled true ### Reported statistics Uncertainty 95% confidence interval: 35.65–51.85% confidenceInterval95 8.10 problems 48 ### Source [Original source](https://matharena.ai/competition_tables/arxiv--june) Source matharena Original model label DeepSeek-v4-Flash (Max) Source version arxiv--june Observed 2026-10-22T05:30:09.097Z Evidence SHA-256 `6832faff9be7e16a6c646c59017e30a2caedebe8cf02caf617a5ed9d8fb03bab` Source locator `/table/rows/26` | | | | | | | | | [BrokenArXiv 06/2026](https://benchize.com/benchmarks/matharena%3Aarxiv_false--june.md) | 16.67% | 95% CI 10.93–22.41% | 21 / 26 | 0.04usd per problem | — | [matharena · arxiv_false--june](https://matharena.ai/competition_tables/arxiv_false--june) | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 16.67 \| percent \| competition \| Higher is better \| \| `cost_per_problem` \| 0.037 \| usd_per_problem \| competition \| Lower is better \| ### Configuration competition arxiv_false--june predicted false Reasoning effort max sourceModelName DeepSeek-v4-Flash (Max) sourceModelVersion deepseek/deepseek_v4_flash thinkingEnabled true ### Reported statistics Uncertainty 95% confidence interval: 10.93–22.41% confidenceInterval95 5.74 problems 54 ### Source [Original source](https://matharena.ai/competition_tables/arxiv_false--june) Source matharena Original model label DeepSeek-v4-Flash (Max) Source version arxiv_false--june Observed 2026-10-22T05:30:06.636Z Evidence SHA-256 `f6d7e2280d05706643d46ae7786bd3fad6d912e4893781c3f62c73deebf38459` Source locator `/table/rows/22` | | | | | | | | | [ArXivMath Overall](https://benchize.com/benchmarks/matharena%3Aoverall--final_answer_comps.md) | 76.55% | 95% CI 73.65–79.45% | 7 / 20 | 0.03usd per problem | — | [matharena · overall--final_answer_comps](https://matharena.ai/competition_tables/overall--final_answer_comps) | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 76.55 \| percent \| competition \| Higher is better \| \| `cost_per_problem` \| 0.027 \| usd_per_problem \| competition \| Lower is better \| ### Configuration competition overall--final_answer_comps predicted false Reasoning effort max sourceModelName DeepSeek-v4-Flash (Max) sourceModelVersion deepseek/deepseek_v4_flash thinkingEnabled true ### Reported statistics Uncertainty 95% confidence interval: 73.65–79.45% confidenceInterval95 2.90 problems Not reported ### Source [Original source](https://matharena.ai/competition_tables/overall--final_answer_comps) Source matharena Original model label DeepSeek-v4-Flash (Max) Source version overall--final_answer_comps Observed 2026-10-22T05:30:11.571Z Evidence SHA-256 `8b08b20c1ac5aec9a829d234469bc39d40b0eb189790d84b7264c2756ffa161e` Source locator `/table/rows/7` | | | | | | | | | [BrokenArXiv 04/2026](https://benchize.com/benchmarks/matharena%3Aarxiv_false--april.md) | 23.77% | 95% CI 16.22–31.32% | 9 / 18 | 0.03usd per problem | — | [matharena · arxiv_false--april](https://matharena.ai/competition_tables/arxiv_false--april) | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 23.77 \| percent \| competition \| Higher is better \| \| `cost_per_problem` \| 0.033 \| usd_per_problem \| competition \| Lower is better \| ### Configuration competition arxiv_false--april predicted false Reasoning effort max sourceModelName DeepSeek-v4-Flash (Max) sourceModelVersion deepseek/deepseek_v4_flash thinkingEnabled true ### Reported statistics Uncertainty 95% confidence interval: 16.22–31.32% confidenceInterval95 7.55 problems 61 ### Source [Original source](https://matharena.ai/competition_tables/arxiv_false--april) Source matharena Original model label DeepSeek-v4-Flash (Max) Source version arxiv_false--april Observed 2026-10-22T05:30:06.112Z Evidence SHA-256 `e543f14dac0ba9663e8104a87738726c9848b89171a53fdca89b7baa4a892399` Source locator `/table/rows/9` | | | | | | | | | [BrokenArXiv 05/2026](https://benchize.com/benchmarks/matharena%3Aarxiv_false--may.md) | 19.5% | 95% CI 11.73–27.27% | 14 / 22 | 0.04usd per problem | — | [matharena · arxiv_false--may](https://matharena.ai/competition_tables/arxiv_false--may) | | | ### Original source values \| Metric \| Value \| Unit \| Scope \| Direction \| \| --- \| --- \| --- \| --- \| --- \| \| `accuracy` \| 19.50 \| percent \| competition \| Higher is better \| \| `cost_per_problem` \| 0.036 \| usd_per_problem \| competition \| Lower is better \| ### Configuration competition arxiv_false--may predicted false Reasoning effort max sourceModelName DeepSeek-v4-Flash (Max) sourceModelVersion deepseek/deepseek_v4_flash thinkingEnabled true ### Reported statistics Uncertainty 95% confidence interval: 11.73–27.27% confidenceInterval95 7.77 problems 50 ### Source [Original source](https://matharena.ai/competition_tables/arxiv_false--may) Source matharena Original model label DeepSeek-v4-Flash (Max) Source version arxiv_false--may Observed 2026-10-22T05:30:06.354Z Evidence SHA-256 `b317fca30109e8c78ed5d2b835fe227b8cfbee3b6f9cb6d7aaeadca866660969` Source locator `/table/rows/14` | | | | | | | | --- Cite as: DeepSeek V4 FlashMax — Benchize, https://benchize.com/models/deepseek-v4-flash-max (retrieved 2026-10-22)