Salamandra 40B Instruct
By BSC · Spain
Updated 2026-07-13
Overview
BSC's 40B scaled-up Salamandra covering 35 EU languages with native Catalan support — though the HuggingFace repo is gated and successor ALIA-40B is now available.
When to pick this model
- EU-sovereign workloads needing 40B-class quality
- Romance-language content generation, especially Catalan
- Public-sector and regulated deployments in Europe
- Multilingual research baselines across 35 EU languages
- Workflows already provisioned for ALIA-40B comparisons
VRAM requirements by quantization
| Quantization | VRAM required |
|---|---|
| Q4_K_M (recommended) | 24 GB |
| Q5_K_M | 29 GB |
| Q8_0 | 43 GB |
| FP16 (no quantization) | 80 GB |
VRAM figures include model weights plus a typical 8k KV cache and ~600 MB runtime overhead (Ollama / llama.cpp baseline). Add headroom for higher context lengths.
In practice, Salamandra 40B Instruct wants a 24 GB card at Q4_K_M (24 GB). Stepping up to Q8_0 nearly doubles the footprint to 43 GB, and unquantized FP16 weights take 80 GB — plan your GPU around the Q4 or Q5 figure unless you specifically need the higher fidelity.
Without a GPU, Salamandra 40B Instruct needs roughly 40 GB of system RAM to run on CPU via llama.cpp or Ollama — workable for background jobs, but far slower than GPU inference. Throughput estimates from our compatibility engine: around 2 tokens/sec on entry-level GPUs, on the order of 10 tokens/sec on a mid-range card, and up to 25 tokens/sec on high-end hardware — assuming the chosen quantization fully fits in VRAM.
What hardware do you need
The table below matches Salamandra 40B Instruct to common GPU memory tiers, using the highest-fidelity quantization that fully fits each card class. Spilling layers to system RAM works but costs most of the speed, so size your card to the quantization you actually want to run.
| GPU memory | Example cards | Best fit for Salamandra 40B Instruct |
|---|---|---|
| 8 GB | RTX 5070 Laptop, RTX 5060, RTX 5060 Ti 8GB | Does not fit — needs 24 GB at Q4_K_M |
| 12 GB | RTX 5070, RTX 5070 Ti Laptop, RTX 4080 Laptop | Does not fit — needs 24 GB at Q4_K_M |
| 16 GB | RTX 5080, RTX 4080 Super, Radeon RX 9070 XT | Does not fit — needs 24 GB at Q4_K_M |
| 24 GB | RTX 4090, Radeon RX 7900 XTX, RTX 5090 Laptop | Q4_K_M (24 GB used) |
| 32 GB | RTX 5090 | Q5_K_M (29 GB used) |
Which GPU should you buy to run Salamandra 40B Instruct?
To run Salamandra 40B Instruct locally at Q4, you need ~24 GB of VRAM. The best value for this is a RTX 4090 (24 GB VRAM).
As an Amazon Associate, BestLLMfor earns from qualifying purchases, at no extra cost to you. It does not influence our independent rankings.
Strengths
- Sovereign European model purpose-built for Romance languages
- Unique native Catalan capability among open models
- Apache 2.0 license
- 7.68T tokens with strong Iberian-language coverage
Limitations
- ~24 GB VRAM at Q4
- 8192-token context limits modern long-context use
- Limited fine-tune ecosystem and gated repo access
Typical workloads
In our catalog grid, Salamandra 40B Instruct is filed under Advanced EU Multilingual — the use cases where its size/quality trade-off makes the most sense. Its tags translate to concrete workloads: multilingual workloads; French-language output where quality matters.
Note the 8k-token context window — fine for short interactions, limiting for long documents or big retrieval contexts. The Apache 2.0 license is permissive, so shipping it inside a commercial product raises no special legal questions.
Architecture & training
Architecture: Dense · 40B · BSC MareNostrum · sovereign Romance languages
Training: Barcelona Supercomputing Center — 7.68T tokens, strong in Catalan, Spanish, French, Occitan.
The strongest open model for Catalan and Iberian Romance languages — but check ALIA-40B first if you can run either.
Quick start
ollama pull hf.co/BSC-LT/salamandra-40b-instruct-GGUFOr use the open-source MCP server to query this model from Claude Desktop, Cursor, or any MCP-compatible client.
Similar models worth comparing
Frequently asked questions
How much VRAM does Salamandra 40B Instruct need?
At the recommended Q4_K_M quantization, Salamandra 40B Instruct needs about 24 GB of VRAM. Q8_0 takes 43 GB, and unquantized FP16 weights take 80 GB.
Can Salamandra 40B Instruct run without a GPU?
Yes — with roughly 40 GB of system RAM it runs CPU-only through llama.cpp or Ollama. Expect a fraction of GPU speed, which is fine for background or batch jobs but slow for interactive chat.
What context window does Salamandra 40B Instruct support?
Salamandra 40B Instruct supports a 8k-token context window (8,192 tokens).
Can I use Salamandra 40B Instruct commercially?
Yes. Salamandra 40B Instruct is released under Apache 2.0, a permissive open-source license that allows commercial use, modification and redistribution.
How fast is Salamandra 40B Instruct on consumer hardware?
Our compatibility engine estimates on the order of 10 tokens/sec on a mid-range GPU and up to 25 tokens/sec on high-end cards, assuming the quantization fully fits in VRAM.
Which quantization of Salamandra 40B Instruct should I download first?
Start with Q4_K_M (24 GB) — the standard size/quality sweet spot. Step up to Q5_K_M or Q8_0 only if you have VRAM headroom. On a 24 GB card you can run up to Q4_K_M.