Interfaze

logo

we're hiring

pricing

help

docs

blog

sign in

All models

Edge0 145B NVFP4

Edge0 145B NVFP4 by Edge0, a text-generation model with multimodal capabilities. Understand and compare multimodal features, benchmarks, and capabilities.

Comparison

FeatureEdge0 145B NVFP4Interfaze
Input Modalities

text, image

image, text, audio, video, document

Native OCRNoYes
Long Document ProcessingNoYes
Language Support

unknown

162+

Native Speech-to-TextNoYes
Native Object DetectionNoYes
Guardrail ControlsNoYes
Context Input Size

256K

1M

Tool CallingYes

Tool calling supported + built in browser, code execution and web search

Scaling

FeatureEdge0 145B NVFP4Interfaze
Scaling

Self-hosted/Provider-hosted with quantization

Unlimited

Try InterfazeRead the Docs

View model card on Hugging Face

Rebuilt, not converted

Edge0 is not a relabelled Qwen3.8-Flash-Next, and not a re-quantization of it: we rebuilt the model's expert layer. In all 48 layers the routed MoE experts were replaced by a latent MoE — the token is projected down to a narrow latent width (ℓ = 1792 on the 145B, 1280 on the 120B), the experts run there, and the result is projected back to 2560. One expert is now a fixed 1.63 MB (145B) or 1.16 MB (120B) record instead of a full-width block. The architecture itself changed, so the model was post-trained afterwards on a long schedule over our own data; it is trained natively at 4 bits, not quantized after the fact.

That is what matters for a device. On-device inference runs out of memory capacity and memory bandwidth long before it runs out of compute, and the expert bytes a token pushes through set both decode speed and how much memory a machine needs:

Qwen3.8-Flash-NextEdge0-145BEdge0-120B
Routed expert parameters120.8 B84.6 B (−30%)60.4 B (−50%)
Expert parameters per token2.36 B1.65 B1.18 B
Expert pool on disk · NVFP4 builds63.2 GiB44.3 GiB31.6 GiB
Decode · 16 GiB VRAM + 20 GiB RAM18.5 tok/s36.8 tok/s62.4 tok/s

On that budget, decode runs 2.0× Qwen's on the 145B and 3.4× on the 120B. Quality holds through the rebuild: the 145B is ahead on the agentic benchmarks — deepswe 1.1 64.6 vs 58.7, terminal-bench 2.1 86.5 vs 83.3 — and level elsewhere.

Capacity controls · VRAM and host RAM

All of these live in the pack's model.json, not on a command line: flags go in args, environment variables in env, with paths relative to the pack directory. That is the same file the Edge0 Chat app and edge0-infer read, so a change there applies to every launch.

"args": ["--expert-cache", "auto", "--kv-resident", "32768", "--vram-reserve-mib", "1024"],
"env":  {"STRATA_LATENT_DENSE": "./latent_dense_q8.bin", "STRATA_LATENT_DENSE_FMT": "8",
         "STRATA_LAT_FUSED_Q3K": "0"}

VRAM

FlagMeaningDefault
--vram-reserve-mib NVRAM left free for the desktop, the driver and other processes; the engine sizes its caches from what remains. It means "leave this much free", not "give the engine this much", so the same number means different things on cards with different totals. Accepted but not listed by --help.700
--expert-cache N|autoExpert slots kept in VRAM. auto derives the count from the free VRAM after the fixed footprint (skeleton projections ≈ 2.9 GiB, MTP draft ≈ 0.8 GiB, verify and draft-head buffers ≈ 0.45 GiB, plus the D/U device buffers).0 (none)
--expert-cache-device1..3 N|autoUse the second and third GPU as further expert-cache tiers.off
--kv-resident NKeep N KV cells per QSA layer in VRAM and stream the K/V from pinned host RAM, which frees VRAM for experts.0 (all in VRAM)
--mtp-q4 all|proj|head4-bit copies of the draft layer's projections and/or head, about 0.3–0.5 GiB less VRAM.off
--prefill CHUNKPrompt chunk size; bigger chunks need bigger device buffers. auto uses the largest chunk whose buffers the expert cache can lend.auto
--spec NThe MTP drafter's verify window: how many tokens it proposes per check. A bigger window needs bigger verify buffers (about 180 MiB at 4) and pays off when every window has a fixed per-layer cost. The shipped packs use 4; the macOS setup picks 2, or 3 when the experts stream from SSD. Without an MTP layer there is no speculation.4

Host RAM

FlagMeaningDefault
--memory-limit-mib NThe RAM this engine's container or cgroup may use, for when the container cannot see its own limit.0 (detect)
--resident-budget-gib NCap on the host-RAM expert cache. It holds only the experts the VRAM cache does not hold, so the two sets are disjoint, and the value is clamped to the RAM actually available minus headroom. Requires a static --expert-profile.0 (off)
--resident-expertsThe same with the whole complement as the budget.off
--mmap-expertsRead experts through the pack file on demand instead of pinning them, so nothing is resident in RAM.off
--conversation-cache-mib NConversation cache. 0 turns it off.adaptive
--ple-io direct|mmap|ramThe PLE table on its SSD: direct reads it unbuffered and it never enters memory; ram locks the whole table in RAM.direct
--kv-resident NThe streamed K/V lives in pinned host RAM (1.55 GiB at 32768 cells).—

Community builds

Third-party conversions of this model. They are not made or endorsed by Edge0; this model's licence and the upstream terms still apply to them.

  • infosave/Edge0-145B-cmf — a single-file CMF build for cortiq (Rust, CUDA and Vulkan, OpenAI-compatible). The routed experts are this NVFP4 build's weights copied bit for bit; the D/U projections and the shared expert are re-encoded to 8 bits and the router to 16. The author reports it running from 12 GB cards.

Want more deterministic results?

Interfaze

logo

Product

Playground

OCR

Models

Leaderboards

Pricing

OpenWebSearch

DefaultModel