Your sovereign, private DeepSeek V4 Flash Official
DeepSeek V4 Flash Official (0731) is the official July release of DeepSeek-V4-Flash — a 284B-parameter MoE (13B active) with a 1M-token context, hybrid compressed-sparse attention, and three reasoning-effort modes (non-think / think-high / think-max). A reasoning model in the medium tier, 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 Flash Official compares
Independent intelligence index for DeepSeek V4 Flash Official against every peer in the Medium tier (and the tier above where needed) — same source, same tests.
Latency — end-to-end seconds for a 500-token answer
Capability indexes
Domain-weighted agentic performance for DeepSeek V4 Flash Official.
Capability index vs medium peers
The evaluations behind the index
Independent benchmark results for DeepSeek V4 Flash Official — Elo scores as published; pass rates shown as percentages.
Per-evaluation economics
| Evaluation | Score | Time / task | Output tokens / task |
|---|---|---|---|
| AA-Briefcase | 1259.67 | 5.7 min | 108330 |
| GDPval-AA | 1426.57 | 5.0 min | 94770 |
| AutomationBench-AA | 54.0% | 1.9 min | 36294 |
| Terminal-Bench 4.0 | 12.1% | 7.8 min | 146812 |
| SciCode | 50.3% | 1.7 min | 31296 |
| Humanity's Last Exam | 38.6% | 1.7 min | 32354 |
| GDP.pdf | 11.0% | 1.3 min | 24138 |
| CritPt | 16.6% | 4.5 min | 85826 |
| AA-Omniscience | -1428.3% | 0.8 min | 15070 |
| AA-LCR (long context) | 79.7% | 0.2 min | 4204 |
Run DeepSeek V4 Flash Official privately
One key, one endpoint, three jurisdictions.
- Size tierMedium
- Serving routeStandard
- Input rate$0.42/1M
- Output rate$1.51/1M
- Cached input rate$0.10/1M
- Context window450K 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