Your sovereign, private DeepSeek V4.1 Flash 1017 Ultra-fast
DeepSeek V4.1 Flash 1017 is the October point-release of DeepSeek-V4.1-Flash: a multimodal MoE with native image+text understanding at up to 1M token context and a Causal Encoder-Decoder architecture that keeps KV-cache footprint roughly a quarter of V4-Flash for cost-efficient agentic workloads. Served on the Fireworks Ultra-fast (Preview) tier: measured ~3.7x the standard output speed (660 vs 172 tok/s) at a premium price. Text and tool calling only on this tier; image input is not served by the Ultra-fast router. A reasoning model in the medium tier on the ultrafast serving route, served on ToothFairyAI's AU, EU and US endpoints — private by default, billed in US dollars per million tokens at the published rates below. No training on your data, no vendor lock-in.
How DeepSeek V4.1 Flash 1017 Ultra-fast compares
Independent intelligence index for DeepSeek V4.1 Flash 1017 Ultra-fast against every peer in the Medium tier (and the tier above where needed) — same source, same tests.
Throughput — tokens per second, median
Latency — end-to-end seconds for a 500-token answer
Capability indexes
Domain-weighted agentic performance for DeepSeek V4.1 Flash 1017 Ultra-fast.
Capability index vs medium peers
The evaluations behind the index
Independent benchmark results for DeepSeek V4.1 Flash 1017 Ultra-fast — Elo scores as published; pass rates shown as percentages.
Per-evaluation economics
| Evaluation | Score | Time / task | Output tokens / task |
|---|---|---|---|
| AA-Briefcase | 1570.42 | 10.8 min | 201620 |
| GDPval-AA | 1745.77 | 7.5 min | 140210 |
| AutomationBench-AA | 66.6% | 3.4 min | 63396 |
| Terminal-Bench 4.0 | 33.8% | 11.7 min | 218213 |
| SciCode | 55.9% | 0.7 min | 12837 |
| Humanity's Last Exam | 43.3% | 2.0 min | 38053 |
| GDP.pdf | 17.2% | 0.9 min | 16372 |
| CritPt | 24.0% | 7.1 min | 131319 |
| AA-Omniscience | 4.70 | 0.1 min | 1439 |
| AA-LCR (long context) | 82.7% | 0.2 min | 3317 |
Run DeepSeek V4.1 Flash 1017 Ultra-fast privately
One key, one endpoint, three jurisdictions.
- Size tierMedium
- Serving routeUltrafast
- Input rate$3.78/1M
- Output rate$13.59/1M
- Cached input rate$0.08/1M
- Context window1049K tokens
- Max output16K tokens
- ReasoningYes — emits thinking traces
- Vision & video inputText only
- Tool callingSupported
- Data residencyAU, EU and US endpoints — processing stays inside the region you call