models · rekaai
Reka Flash 3 — catalogued from third-party evaluation data; no weights held by Sovereign Frontier.
| Method | hf-published-sha256 |
|---|
| Publisher | rekaai |
|---|---|
| Origin | United States |
| Licence | Apache-2.0 |
| Parameters | 20,905,482,240 (20.9B) |
| Context window | 65,536 tokens |
| Modality | text->text |
| Architecture | llama |
| Layers | 44 |
| Hidden size | 6144 |
| Attention | GQA (64 heads, 8 KV) |
| Native precision | BF16 |
| Pulls recorded here | 7 |
| Downloads reported upstream | 614 |
| MMLU-Pro | 66.9 |
|---|---|
| GPQA Diamond | 52.9 |
| AIME 2025 | 33.7 |
| MATH-500 | 89.3 |
| LiveCodeBench | 43.5 |
| IFBench | 30.4 |
| Humanity's Last Exam | 4.4 |
Reported by a third party and recorded here, not re-run by us. Source: artificialanalysis.ai/api/v2.
| Hardware | Engine | Build measured | Decode, 1 stream | Label | Reported by |
|---|---|---|---|---|---|
| GeForce RTX 3090 | llama.cpp | Q4_K_M | 36.3 tok/s @ 512 | candidate | localmaxxing |
| GeForce RTX 3060 ×2 devices | llama.cpp | Q4_K_M | 23.9 tok/s @ 512 | candidate | localmaxxing |
From local-ai-registry (MIT), commit 124959d: 2 runs across 2 machines, 0 validated — model revision and runtime pinned, launch accepted — and 2 candidate, their label for useful evidence without a reproducible-launch promise.
Registry record as of 2026-08-13. Identity, licence and parameter count are properties of a release and do not move; tier can, and the verify link above re-checks it against the live registry rather than this page.