DeepSeek V4.1 Flash 1017 Is Live on Multos — The October Snapshot That Outcodes Its Own Family
The 1017 snapshot of DeepSeek V4.1 Flash (October 17, 2026) lifts HLE from 36.8 to 42.6 and Terminal-Bench 4.0 from 31.2 to 42.3 over the base Flash — at the same Lite-tier price. Live on Multos via Novita and Baseten.
On October 17, 2026, DeepSeek shipped a snapshot update to V4.1 Flash — and the numbers in the model card read like a new generation, not a point release. HLE without tools jumps from 36.8 to 42.6. Terminal-Bench 4.0 goes from 31.2 to 42.3. DeepSWE v1.1 hits 77.0. The 1017 snapshot is now live on Multos Lite and above, served through Novita and Baseten.
What the 1017 Snapshot Actually Is
DeepSeek-V4.1-Flash-1017 is the October 17 snapshot of V4.1 Flash, DeepSeek's cost-efficient tier. Architecturally it is a sparse mixture-of-experts model built on the Causal Encoder-Decoder (CED) architecture — the first DeepSeek model on it — with a 552B-parameter backbone and an asymmetric activation split: 8 billion parameters on input, 16 billion on output. Vision is native, with visual and text embeddings trained jointly from the start of pre-training rather than bolted on afterward. Compressed KV caching cuts cache memory to roughly a quarter of the previous Flash generation, which is a direct cost saving on long agentic runs. The weights are MIT licensed.
Benchmarks: The Gains Are Real
From DeepSeek's model card, 1017 versus the base V4.1 Flash: HLE without tools 42.6 versus 36.8; Terminal-Bench 4.0 42.3 versus 31.2 (Claude Code harness); DeepSWE v1.1 77.0 versus 74.2; ProgramBench 29.1 versus 20.3; Automation-Bench v1.0.6 60.5 versus 54.8; Agents' Last Exam 35.4 versus 31.8; Apex Bench v1.1 72.1 versus 64.6. At reporting time that Terminal-Bench number ranked second among models tracked on the HF eval tracker. Two honest notes. First, harness choice moves these numbers — Artificial Analysis, running its own harness at max reasoning, posts a lower Terminal-Bench score — so compare like with like. Second, the gains are not just quality: DeepSeek reports reduced task execution time and shorter outputs, which lowers the cost of completing a task, not just the cost per token.
What This Means for Building on Multos
The 1017 snapshot is live on Multos Lite and above through Novita and Baseten. Throughput on those providers runs around 90-105 output tokens per second, and the compressed KV cache keeps long sessions cheap. It is part of the Lite fallback chain, so Auto Router and the failover path can land on it when a primary model stumbles. The fit: high-volume coding, terminal and computer-use agents, long-horizon tasks that must run to completion, and long-context analysis — at Lite-tier pricing, with vision included.
Where to Find It
Find DeepSeek V4.1 Flash 1017 in the Novita AI and Baseten groups of the model dropdown, available on Lite and above.
Ready to Get Started?
DeepSeek V4.1 Flash 1017 is live on Multos Lite and above — the October snapshot that lifts HLE by 6 points and Terminal-Bench by 11 over its own base, at the same Lite-tier price. Pick it from the Novita AI or Baseten group or let Auto Router route to it.
Start Building FreeRelated Articles
Claude Opus 5.5 Is Live on Multos — Fable-Class Work at 40% Lower Cost
Anthropic released Claude Opus 5.5 today and it's immediately available on Multos Elite. It matches Claude Fable 5.1 on most tasks, costs 40% less to run than Opus 5, and is 30% faster. On Terminal-Bench 4.0 it scores 66.4% — higher than GPT-6 Astra, higher than Opus 5, higher than Fable 5.1. Here's what that actually means for building apps and games on Multos.
Read more →GPT-6 Sol and GPT-6 Luna Are Live — OpenAI Halves the Price of Its Best Models
OpenAI released GPT-6 Sol and GPT-6 Luna today, cutting prices by 50% versus GPT-5.6. Sol drops from $4/$20 to $2/$10 per million tokens. Luna drops from $0.20/$1.20 to $0.10/$0.50. Intelligence Index scores hold steady, hallucinations drop significantly, and both models are live on Multos now — Sol on Pro, Luna on Lite.
Read more →MiMo V2.6 Flash and Pro Are Now on Multos — Frontier-Grade Agents at Flash-Tier Pricing
Xiaomi released MiMo V2.6 Pro and Flash on September 21, 2026. Pro scored 46 on the Artificial Analysis Intelligence Index — the highest any open-weight model has ever scored, at $0.13 per task. Flash costs $0.14/$0.28 per million tokens and trails Pro by only 4 points on DeepSWE. Both are live on Multos Lite via DeepInfra.
Read more →