mxbai Embed Large needs ~8.2 GB VRAM. NVIDIA L20 48GB has 48.0 GB. With F16 quantization, expect ~5 tok/s.
Operating mode
Interactive favors responsiveness, while light API and scale-out lean harder on serving readiness. The fit stays the same, but the recommendation lens changes.
Current mode
Balanced
Balanced for general local use. Keeps the ranking neutral across personal and serving workflows.
Select quantization to explore
Fit status
Runs well
Decode
4.7 tok/s
TTFT
41279 ms
Safe context
512
Memory
8.2 GB / 48.0 GB
This model fits, but memory bandwidth is the part holding decode speed back.
Throughput will feel slow
Estimated decode speed is only 4.7 tok/s, so this is more of a technical fit than a comfortable daily-driver setup.
Prioritize bandwidth, not only capacity
If this workload feels slow, the next useful step is often a GPU tier with materially faster memory bandwidth rather than only a small bump in capacity.
| Workload | Grade | Fit | Decode | TTFT | Context |
|---|---|---|---|---|---|
| Chat | A | Runs well | 4.7 tok/s | 22516 ms | 512 |
| Coding | A | Runs well | 4.7 tok/s | 41279 ms | 512 |
| Agentic Coding | A | Runs well | 4.7 tok/s | 60043 ms | 512 |
| Reasoning | A | Runs well | 4.7 tok/s | 48785 ms | 512 |
| RAG | A | Runs well | 4.7 tok/s | 75053 ms | 512 |
Inference speed
Estimated decode speed (tokens/sec) for mxbai Embed Large at F16 across popular GPUs and Apple Silicon, using the fastest local runtime per device. Fastest is RTX 5090 32GB at ~5 tok/s. Speed is memory-bandwidth bound, so cards that fit the whole model in VRAM run far faster than ones that offload to system RAM.
| GPU / Mac | Memory | Quant | Speed (tok/s) | Fits? |
|---|---|---|---|---|
 RTX 5090 32GB | 32 GB | F16 | 4.7 | Fits |
| RTX 4090 24GB | 24 GB | F16 | 4.7 | Fits |
| RTX 4080 Super 16GB | 16 GB | F16 | 4.7 |
How mxbai Embed Large (0.33500000834465027B params) fits at each quantization level on NVIDIA L20 48GB (48.0 GB usable).
| Quant | Bits | VRAM | Quality | Fit |
|---|---|---|---|---|
Q2_K | 2 | 0.1 GB | Low | A75 |
Q3_K_S | 3 | 0.2 GB | Low | A75 |
NVFP4 | 4 |
Copy-paste commands to run mxbai Embed Large on your machine.
Run
ollama run mxbai-embed-largeYour hardware
| Model | Params | Grade | Decode | Capabilities |
|---|---|---|---|---|
 Qwen3-Coder 30B A3B Instruct | 30.5B | S | 95.4 tok/s | |
| Qwen 3.5 27B | 27B | S | 41.4 tok/s |
Yes, NVIDIA L20 48GB can run mxbai Embed Large with a A grade (Runs well). Expected decode speed: 4.7 tok/s.
mxbai Embed Large (0.33500000834465027B parameters) requires approximately 8.2 GB of memory with F16 quantization.
The recommended quantization for mxbai Embed Large is F16, which balances quality and memory efficiency.
On NVIDIA L20 48GB, mxbai Embed Large achieves approximately 4.7 tokens per second decode speed with a time-to-first-token of 41279ms using F16 quantization.
For coding workloads, mxbai Embed Large on NVIDIA L20 48GB receives a A grade with 4.7 tok/s and 512 context.
On NVIDIA L20 48GB, mxbai Embed Large can safely use up to 512 tokens of context. The model's official context limit is 512, but available memory constrains the safe maximum.
Paste this snippet into any page to show a live fit card.
<iframe src="https://willitrunai.com/embed/mxbai-embed-large-on-l20-48gb" width="400" height="180" frameborder="0" style="border:none;border-radius:12px;overflow:hidden;" title="Will It Run AI — fit result"></iframe>
Preview:
| RTX 3090 24GB | 24 GB | F16 | 4.7 | Fits |
| RTX 4070 12GB | 12 GB | F16 | 4.7 | Fits |
| RTX 3060 12GB | 12 GB | F16 | 4.7 | Fits |
| RTX 4060 8GB | 8 GB | F16 | 4.7 | Fits |
RX 7900 XTX 24GB | 24 GB | F16 | 4.7 | Fits |
MacBook Pro M4 Max 128GB | 128 GB | F16 | 4.7 | Fits |
Mac Studio M3 Ultra 256GB | 256 GB | F16 | 4.7 | Fits |
Mac Studio M2 Ultra 128GB | 128 GB | F16 | 4.7 | Fits |
Mac Studio M1 Ultra 128GB | 128 GB | F16 | 4.7 | Fits |
MacBook Pro M4 Max 64GB | 64 GB | F16 | 4.7 | Fits |
MacBook Pro M3 Max 64GB | 64 GB | F16 | 4.7 | Fits |
MacBook Pro M1 Max 64GB | 64 GB | F16 | 4.7 | Fits |
MacBook Pro M4 Pro 48GB | 48 GB | F16 | 4.7 | Fits |
Estimates for single-stream decoding at F16; real tokens/sec varies with prompt length, context, batch size, and runtime build. Prompt processing (prefill) is faster than the decode figures shown here.
0.2 GB |
| Medium |
| A75 |
Q4_K_M | 4 | 0.2 GB | Medium | A75 |
Q5_K_M | 5 | 0.2 GB | High | A75 |
Q6_K | 6 | 0.3 GB | High | A75 |
Q8_0 | 8 | 0.4 GB | Very High | A75 |
F16Best for your GPU | 16 | 0.7 GB | Maximum | A75 |
| Qwen 3.6 27B | 27B | S | 41.5 tok/s |
| Qwen 3.6 35B A3B | 35B | S | 85.8 tok/s |
| Qwen3-VL 30B A3B Instruct | 30B | S | 98.6 tok/s |
Prioritize bandwidth, not only capacity. If this workload feels slow, the next useful step is often a GPU tier with materially faster memory bandwidth rather than only a small bump in capacity.