DeepSeek
DeepSeek V4.1 Flash-1017
The A-tier cost-conscious coding option, represented by DeepSeek’s latest Flash-1017 visual run.
API pricing
USD per million tokens · DeepSeek API peak rate- Input
- $0.3
- Output
- $1.2
- Cached input
- $0.006
- Context window
- 1M
Try the active visual tests
Open the generated scene to test it, or compare that prompt with another model.

Last-Light Relay
Run date: 2026-10-26
Generation provenance: unrecorded
Superbench 2.0 brief · one-shot authoring, no repair turn
Resource-constrained scheduling testing exact optimality, frozen-history recovery, and replanning after a failure.

Prism Freight Pavilion
Run date: 2026-10-26
Generation provenance: unrecorded
Superbench 2.0 brief · one-shot authoring, no repair turn
Single-message exhibit assignment testing exact enumeration, infeasibility explanation, and a shared editable state.

Quarantine Signal Desk
Run date: 2026-10-26
Generation provenance: unrecorded
Superbench 2.0 brief · one-shot authoring, no repair turn
Trust-boundary screening desk testing policy precedence, atomic review steps, and hostile text handled strictly as data.

Tidal Ledger Observatory
Run date: 2026-10-26
Generation provenance: unrecorded
Superbench 2.0 brief · one-shot authoring, no repair turn
Event-sourced ledger testing deterministic replay, idempotent imports, and timeline branching under edited data.
Archived model versions
The latest DeepSeek V4.1 Flash-1017 runs are shown above. Earlier versions are preserved separately and are not counted in this current run set.
Local benchmark run
Four Superbench 2.0 briefs, one-shot authoring
Sources, pricing details and methodology
Run identity
Model specifications
- Status
- Available in this Codex session
- Codex model ID
deepseek-flash- Run date
- October 26, 2026
- Access route
- DeepSeek API alias deepseek-flash. DeepSeek maps that alias to DeepSeek-V4.1-Flash-1017 as of the October 17, 2026 update, so pinning the alias does not freeze the served version. Record the served version where the provider exposes it.
Flash-1017 keeps the September V4.1 Flash list prices. Peak hours are Monday-Friday, 01:00-04:00 and 06:00-10:00 UTC; off-peak rates are half: $0.15 input, $0.003 cached input, and $0.60 output per 1M tokens.
Benchmark sources
Four frozen Superbench 2.0 briefs were assigned to separate DeepSeek Harness subagents on the deepseek-flash alias. Each agent was scoped to one prompt file and one output path, with no reference implementation, other model output or site source, and authored its artifact without browser testing or repair turns.
DeepSeek V4.1 Flash-1017 Superbench 2.0 run recordThese are inspectable local artifacts, not provider benchmark scores or an aggregate rating. The run record states which source-level checks were performed, which checks remain for a human browser review, and where the delivery differs from a hosted-API single-message episode. It does not claim the briefs were passed.
Editorial verdict
Where this model fits
DeepSeek V4.1 Flash-1017 is the October 17 served version behind the deepseek-flash API alias: a 552B-parameter multimodal MoE with a one-million-token context window. Its current Superbash entry contains four Superbench 2.0 artifacts published on October 26. The September V4.1 Flash suite and older V4 Pro and V4 Flash runs remain available as archived evidence, not duplicate current models or transferred scores.
Best for
Watch out
Why it is ranked here
Evidence
Superbash editorial model ranking
Takeaway: DeepSeek V4.1 Flash-1017 is currently placed in Tier A.
The November 2026 editorial roster places DeepSeek V4.1 Flash-1017 at rank 6.
Open source →DeepSeek-V4.1-Flash-1017 release
Takeaway: Flash-1017 is the latest recorded served version behind DeepSeek’s deepseek-flash alias.
The official release establishes the model identity and alias transition; older model scores are not silently reassigned to it.
Open source →Superbench 2.0: DeepSeek V4.1 Flash-1017 run record
Takeaway: Four latest-version visual artifacts are published with their methodology and protocol limits.
The local run record keeps artifact inspection, source-level checks, and unreviewed browser behavior distinct.
Open source →Related guides