deepseek-flash
A lightweight hosted model tier offered by DeepSeek, intended for fast inference and cost-sensitive scenarios, with token-based billing.
Core Facts
This update adds support for the official model ID deepseek-flash; new sessions default to V4.1 Flash, but explicitly configured models take higher priority.
DeepSeek replaced the model behind the deepseek-flash API endpoint with a new checkpoint, DeepSeek-V4.1-Flash-1017, without a changelog or email notification.
DeepSeek released the V4.1-Flash-1017 model on October 17, adding the Search and Fetch APIs.
Timeline (last 90 days)
On October 17, 2026, DeepSeek replaced the checkpoint behind the deepseek-flash alias with V4.1 Flash 1017, keeping the model name and price unchanged.
DeepSeek replaced the model behind the deepseek-flash API endpoint with a new checkpoint, DeepSeek-V4.1-Flash-1017, without a changelog or email notification.
Silently swapping the model while keeping the model name unchanged means that earlier evaluations run against the deepseek-flash endpoint describe a checkpoint that no longer serves traffic.
DeepSeek released the V4.1-Flash-1017 model on October 17, adding the Search and Fetch APIs.
Tested with deepseek-flash, the classifier feature (high threshold) blocked 218 of 218 MCPTox tools with 0 false positives on 45 clean servers; on 200 random real servers it blocked or flagged from 2 up to 5, and on the 996 sample packages from 23 up to 35.
On accuracy: τ=0.4 is 0.7305, the hosted single run is 0.7085, the local single run is 0.7320, and τ=0.6 is 0.7115.
The workload measured here is the public typed-decisions row set: 400 cases and 2000 decisions, with the local model being Phocinae-Largha-150M-v1 (CPU, fp32) and the hosted model used for comparison being deepseek-flash in the flash tier (temperature 0).
The hosted-side usage across the 400 cases totals 1,222,211 tokens (290,606 input, 931,605 output), i.e. 3,055.5 tokens per case.
In the hybrid mode at τ=0.4, 27 of the 400 cases are escalated (6.8%), billed Tokens drop from 1,222,211 to 82,499, and the saving including plugin overhead is 86.1%.
Calculated at the paired public unit prices ($0.28 per million input, $1.10 per million output), the 400-case single hosted run costs $1.1061, and τ=0.4 costs $0.0747.
3 more timeline events
All Facts (12)
This update adds support for the official model ID deepseek-flash; new sessions default to V4.1 Flash, but explicitly configured models take higher priority.
65%UnverifiedSilently swapping the model while keeping the model name unchanged means that earlier evaluations run against the deepseek-flash endpoint describe a checkpoint that no longer serves traffic.
50%UnverifiedTested with deepseek-flash, the classifier feature (high threshold) blocked 218 of 218 MCPTox tools with 0 false positives on 45 clean servers; on 200 random real servers it blocked or flagged from 2 up to 5, and on the 996 sample packages from 23 up to 35.
50%UnverifiedOn accuracy: τ=0.4 is 0.7305, the hosted single run is 0.7085, the local single run is 0.7320, and τ=0.6 is 0.7115.
50%UnverifiedIn the hybrid mode at τ=0.4, 27 of the 400 cases are escalated (6.8%), billed Tokens drop from 1,222,211 to 82,499, and the saving including plugin overhead is 86.1%.
50%UnverifiedThe workload measured here is the public typed-decisions row set: 400 cases and 2000 decisions, with the local model being Phocinae-Largha-150M-v1 (CPU, fp32) and the hosted model used for comparison being deepseek-flash in the flash tier (temperature 0).
50%UnverifiedThe hosted-side usage across the 400 cases totals 1,222,211 tokens (290,606 input, 931,605 output), i.e. 3,055.5 tokens per case.
50%UnverifiedDeepSeek officially stated that DeepSeek-V4.1-Flash is the smallest model in its new architecture family.
50%Partially VerifiedDeepSeek replaced the model behind the deepseek-flash API endpoint with a new checkpoint, DeepSeek-V4.1-Flash-1017, without a changelog or email notification.
65%Partially VerifiedDeepSeek released the V4.1-Flash-1017 model on October 17, adding the Search and Fetch APIs.
65%UnverifiedOn October 17, 2026, DeepSeek replaced the checkpoint behind the deepseek-flash alias with V4.1 Flash 1017, keeping the model name and price unchanged.
60%UnverifiedDeepSeek officially announced that it will restructure its API billing into a "peak-valley pricing" model on August 17.
50%