Helium 1 2B
By Kyutai · France
Updated 2026-07-13
Overview
Kyutai's 2B multilingual base covering all 24 EU languages, distilled from Gemma 2 — which means Gemma Terms apply on top of CC-BY-SA. Beats Qwen 2.5 1.5B, Gemma 2B, and Llama 3.2 3B at its scale.
When to pick this model
- You need a small multilingual base for fine-tuning across EU languages
- You're building edge or embedded deployments with French as a priority
- You want a European base model with strong sub-3B performance
- You're doing pre-training research and need a clean small foundation
VRAM requirements by quantization
| Quantization | VRAM required |
|---|---|
| Q4_K_M (recommended) | 1.5 GB |
| Q5_K_M | 2 GB |
| Q8_0 | 3 GB |
| FP16 (no quantization) | 5 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, Helium 1 2B fits an 8 GB consumer card at Q4_K_M (1.5 GB). Stepping up to Q8_0 nearly doubles the footprint to 3 GB, and unquantized FP16 weights take 5 GB — plan your GPU around the Q4 or Q5 figure unless you specifically need the higher fidelity.
Without a GPU, Helium 1 2B needs roughly 4 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 35 tokens/sec on entry-level GPUs, on the order of 100 tokens/sec on a mid-range card, and up to 200 tokens/sec on high-end hardware — assuming the chosen quantization fully fits in VRAM.
What hardware do you need
The table below matches Helium 1 2B 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 Helium 1 2B |
|---|---|---|
| 8 GB | RTX 5070 Laptop, RTX 5060, RTX 5060 Ti 8GB | FP16 (5 GB used) |
| 12 GB | RTX 5070, RTX 5070 Ti Laptop, RTX 4080 Laptop | FP16 (5 GB used) |
| 16 GB | RTX 5080, RTX 4080 Super, Radeon RX 9070 XT | FP16 (5 GB used) |
| 24 GB | RTX 4090, Radeon RX 7900 XTX, RTX 5090 Laptop | FP16 (5 GB used) |
| 32 GB | RTX 5090 | FP16 (5 GB used) |
Which GPU should you buy to run Helium 1 2B?
To run Helium 1 2B locally at Q4, you need ~1.5 GB of VRAM. The best value for this is a RTX 5060 (8 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
- Compact multilingual base from a European lab
- Covers all 24 EU languages
- Beats Qwen 2.5 1.5B, Gemma 2B, and Llama 3.2 3B at its scale
- Built by Kyutai
Limitations
- CC-BY-SA 4.0 plus Gemma Terms via distillation
- Base model — not instruction-tuned
- No official Ollama support
Typical workloads
In our catalog grid, Helium 1 2B is filed under Compact EU Multilingual, Fine-tune Base — 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 4k-token context window — fine for short interactions, limiting for long documents or big retrieval contexts. It ships under the CC-BY-SA 4.0 license — commercial use is generally possible but read the specific terms before embedding it in a product.
Architecture & training
Architecture: Dense · GQA · RoPE · distilled from Gemma 2
Training: 2.5T tokens, 24 EU languages.
A strong European small base for fine-tuning — just budget for the dual-license obligations.
Quick start
# HuggingFace : kyutai/helium-1-2bOr 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 Helium 1 2B need?
At the recommended Q4_K_M quantization, Helium 1 2B needs about 1.5 GB of VRAM. Q8_0 takes 3 GB, and unquantized FP16 weights take 5 GB.
Can Helium 1 2B run without a GPU?
Yes — with roughly 4 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 Helium 1 2B support?
Helium 1 2B supports a 4k-token context window (4,096 tokens).
Can I use Helium 1 2B commercially?
Helium 1 2B ships under the CC-BY-SA 4.0 license. Commercial use is generally permitted subject to its terms — review the license text before shipping a product.
How fast is Helium 1 2B on consumer hardware?
Our compatibility engine estimates on the order of 100 tokens/sec on a mid-range GPU and up to 200 tokens/sec on high-end cards, assuming the quantization fully fits in VRAM.
Which quantization of Helium 1 2B should I download first?
Start with Q4_K_M (1.5 GB) — the standard size/quality sweet spot. Step up to Q5_K_M or Q8_0 only if you have VRAM headroom. It already fits an 8 GB card at FP16.