DeepSeek V4 Pro 0813
A large language model released by DeepSeek, billed purely by token via API, providing relatively high performance for programming and general reasoning scenarios.
Core Facts
V4.1-Flash-1017 scores 77.0 on the DeepSWE benchmark.
Artificial Analysis evaluation shows DeepSeek V4.1 Flash scored 40 on the Intelligence Index, surpassing V4 Pro to become DeepSeek's new flagship, featuring native multimodal input, 8B activated parameters on input and 16B on output, released under the MIT license.
Timeline (last 90 days)
When Artificial Analysis ran its intelligence index at the maximum effort level, the September version generated about 250 million output tokens versus a median of 130 million, 62% more than DeepSeek V4 Pro 0813.
V4.1 Flash 1017 was released on October 17, and V4 Pro 0813 was released on August 13.
On the Terminal-Bench 4.0 benchmark, V4.1 Flash 1017 scores 42.3, and V4 Pro 0813 scores 12.4.
Third-party platform AIHubMix list prices: Flash 1017 is $0.3098 input / $1.2392 output, and Pro 0813 is $1.3944 input / $4.1832 output.
Under an assumed Agent bug-fixing session of 40 calls, 1.2 million input tokens (90% cache hit rate) and 60,000 output tokens, calculated at AIHubMix list prices, Flash 1017 costs $0.118 at peak hours and $0.059 off-peak; Pro 0813 costs $0.469 at peak and $0.233 off-peak; Pro costs about four times as much per session as Flash.
Artificial Analysis found that the September version of Flash used 1.62 times as many output tokens on its index as Pro 0813.
On the four benchmarks ProgramBench, NL2Repo-Bench, Automation-Bench and Agents' Last Exam, V4.1 Flash 1017 scores 29.1, 68.2, 60.5 and 35.4 respectively, all higher than V4 Pro 0813's 15.5, 61.5, 43.2 and 25.7.
API calls based on AIHubMix require AIHUBMIX_API_KEY, the base_url is https://aihubmix.com/v1, and the model IDs are deepseek-v4.1-flash-1017 and deepseek-v4-pro-0813 respectively.
On the text-only HLE subset, V4.1 Flash 1017 scores 44.0, surpassing V4 Pro 0813's 42.7.
DeepSeek has not published V4.1 Flash 1017's score on GPQA Diamond; Pro 0813 scores 92.4 on this item, and the September version of Flash scores 90.9.
2 more timeline events
All Facts (12)
V4.1-Flash-1017 scores 77.0 on the DeepSWE benchmark.
65%VerifiedArtificial Analysis evaluation shows DeepSeek V4.1 Flash scored 40 on the Intelligence Index, surpassing V4 Pro to become DeepSeek's new flagship, featuring native multimodal input, 8B activated parameters on input and 16B on output, released under the MIT license.
65%UnverifiedWhen Artificial Analysis ran its intelligence index at the maximum effort level, the September version generated about 250 million output tokens versus a median of 130 million, 62% more than DeepSeek V4 Pro 0813.
50%UnverifiedOn the Terminal-Bench 4.0 benchmark, V4.1 Flash 1017 scores 42.3, and V4 Pro 0813 scores 12.4.
50%UnverifiedThird-party platform AIHubMix list prices: Flash 1017 is $0.3098 input / $1.2392 output, and Pro 0813 is $1.3944 input / $4.1832 output.
50%UnverifiedUnder an assumed Agent bug-fixing session of 40 calls, 1.2 million input tokens (90% cache hit rate) and 60,000 output tokens, calculated at AIHubMix list prices, Flash 1017 costs $0.118 at peak hours and $0.059 off-peak; Pro 0813 costs $0.469 at peak and $0.233 off-peak; Pro costs about four times as much per session as Flash.
50%UnverifiedArtificial Analysis found that the September version of Flash used 1.62 times as many output tokens on its index as Pro 0813.
50%UnverifiedOn the text-only HLE subset, V4.1 Flash 1017 scores 44.0, surpassing V4 Pro 0813's 42.7.
50%UnverifiedDeepSeek has not published V4.1 Flash 1017's score on GPQA Diamond; Pro 0813 scores 92.4 on this item, and the September version of Flash scores 90.9.
50%UnverifiedOn the four benchmarks ProgramBench, NL2Repo-Bench, Automation-Bench and Agents' Last Exam, V4.1 Flash 1017 scores 29.1, 68.2, 60.5 and 35.4 respectively, all higher than V4 Pro 0813's 15.5, 61.5, 43.2 and 25.7.
50%UnverifiedAPI calls based on AIHubMix require AIHUBMIX_API_KEY, the base_url is https://aihubmix.com/v1, and the model IDs are deepseek-v4.1-flash-1017 and deepseek-v4-pro-0813 respectively.
50%UnverifiedV4.1 Flash 1017 was released on October 17, and V4 Pro 0813 was released on August 13.
50%