DeepSeek's 1017 revision stores 890 bytes per token, hallucination rate reaches 67%
Independent testing shows the model answers 43% of Omniscience questions correctly while declining to answer 67% of questions it should have refused.
By Tunde Bailey ·
DeepSeek has released a new revision of its open-weight V4.1 Flash model, labeled 1017, cutting its KV cache requirement sharply while independent testing shows a persistent hallucination rate.
The model is a Mixture-of-Experts system with 552 billion total parameters, 16 billion active during a run. It accepts text and images and supports a 1 million token context window. DeepSeek's product page carries the slogan "wish-driven programming made real."
Compared with V4 Flash, the 1017 revision requires one-quarter of the HBM and one-eighth of the SSD storage for its KV cache, using 890 bytes per token entry. DeepSeek's own benchmarks place the model between GPT-5.6 Sol and Claude Opus 5 on task tests, and above Anthropic's model in those tests.
DeepSeek's cost-per-task chart is built on DeepSWE's September 2 dataset. Both the standard and 1017 variants land in the favorable top-left region, and low-effort runs of the 1017 version complete tasks at a rate comparable to the max settings of costlier competitors.
At DeepSeek's off-peak pricing, input costs $0.15 per million tokens, output costs $0.60, and cache-hit input falls to $0.003. DeepSeek's search feature costs $0.75 per 1,000 queries; the closest competitor, Parallel, charges $1.
Artificial Analysis ranks the 1017 model (max) second among large open-weight models on its intelligence index with a score of 45, against 58 for Claude Opus 5.5 (max with fallback). The model is among the fastest in its category, and only Xiaomi's MiMo V2.6-Pro beats it on cost per task.
On Artificial Analysis' Omniscience benchmark, the model answered 43% of questions correctly, among the best for open-weight models but behind proprietary systems. It hallucinated on 67% of questions it should have declined to answer. The benchmark's main index scores the 1017 model at 5, against 20 for Kimi K3-max and 42 or more for frontier models.
The source article's author notes that Kimi K3, launched several months ago, drew praise as cheap and effective, but the author's own use ended after short fact lookups produced invented information; Artificial Analysis data appears consistent with that account.
Tasks with verifiable outputs, including programming, OCR with review, and format conversion, are suited to the 1017 model because an agent can check its work and retry. Open-ended research, fact lookups, and drafting or interpreting precise documents are better handled by a frontier model or a human.
