Olala v1.0 mini 7A1B
Licence: Apache 2.0 — plain open licence, no revenue or user-count trigger.
OVHcloud's first own LLM, weights published Nov 19, 2026 on Hugging Face (final, base, SFT, summarise and translate repositories, plus GGUF and MLX routes): 6.8B total parameters by the model card's own figures — the launch material's 7B is the maker's rounding — with about 1B active per token, 36 layers (29 Mamba-3 state-space layers and 7 attention layers, so only seven layers keep a KV cache), each followed by an MoE block with 256 routed experts of which 6 fire, a 32,768-token context (a 128K version is promised), and training on 12.8T tokens. Memory by arithmetic: about 13.6 GB at BF16; the GGUF route runs in 6–14GB by the card. Watch-out: upstream vLLM, llama.cpp, Ollama and LM Studio cannot load the olala architecture yet — self-hosting uses OVHcloud's forks (vLLM fork / pre-built Docker image, tested on L4 and L40S; llama.cpp fork for GGUF). All performance figures are OVHcloud's own.
What it needs
How we work this out: memory = the model's total parameters × bytes per weight — 2 bytes at full precision (FP16), 1 byte at 8-bit, half a byte at 4-bit. Weights only — add room for the conversation memory (KV cache) and the serving software. A GPU counts as running a build when its memory covers the weights plus about 20% headroom. Real figures move with context length and software.
GPUs that can run it
Ranked by how well the model runs, then by price.
Runs Olala v1.0 mini 7A1B at full precision — its 14GB of weights fit this card's 24GB, with room left for the conversation memory.
24GB fills up fast on big models. No NVLink, so two cards can't pool memory — and it is a gaming card at heart.
Runs Olala v1.0 mini 7A1B at full precision — its 14GB of weights fit this card's 48GB, with room left for the conversation memory.
An older generation — no FP8, so the newer speed tricks for serving don't apply.
Runs Olala v1.0 mini 7A1B at full precision — its 14GB of weights fit this card's 32GB, with room left for the conversation memory.
No NVLink, so two cards can't pool memory. A consumer card — driver and ECC behaviour matter for production serving.
Runs Olala v1.0 mini 7A1B at full precision — its 14GB of weights fit this card's 95GB, with room left for the conversation memory.
This is the Max-Q version: it runs at lower power and is slower than the full card.
Runs Olala v1.0 mini 7A1B at full precision — its 14GB of weights fit this card's 48GB, with room left for the conversation memory.
70B at FP16 does not fit, and no flag fixes that. Weak at double-precision maths; no NVLink.
Runs Olala v1.0 mini 7A1B at full precision — its 14GB of weights fit this card's 48GB, with room left for the conversation memory.
Weak at double-precision maths — not the card for scientific computing. No NVLink.
Runs Olala v1.0 mini 7A1B at full precision — its 14GB of weights fit this card's 80GB, with room left for the conversation memory.
An older generation — no FP8, so the newer speed tricks for serving don't apply.
Runs Olala v1.0 mini 7A1B at full precision — its 14GB of weights fit this card's 96GB, with room left for the conversation memory.
No NVLink scale-out; strong all-round, but the Max-Q version (listed separately) runs slower.
Runs Olala v1.0 mini 7A1B at full precision — its 14GB of weights fit this card's 40GB, with room left for the conversation memory.
The 40GB fills up quickly on larger models.
Runs Olala v1.0 mini 7A1B at full precision — its 14GB of weights fit this card's 80GB, with room left for the conversation memory.
No NVLink, and lower power than the SXM — slower when cards must work together.
Runs Olala v1.0 mini 7A1B at full precision — its 14GB of weights fit this card's 80GB, with room left for the conversation memory.
Pays off only when the card stays busy; on bursty work you pay H100 rent for idle silicon.
Runs Olala v1.0 mini 7A1B at full precision — its 14GB of weights fit this card's 94GB, with room left for the conversation memory.
Sold as linked pairs, not as a lone card.
Runs Olala v1.0 mini 7A1B at full precision — its 14GB of weights fit this card's 80GB, with room left for the conversation memory.
One provider's label, published alongside its plainer “H100 PCIe” — check which card you are actually booking.
Runs Olala v1.0 mini 7A1B at full precision — its 14GB of weights fit this card's 141GB, with room left for the conversation memory.
More card than most jobs need; the main trap is paying for it on small work.
Runs Olala v1.0 mini 7A1B at full precision — its 14GB of weights fit this card's 80GB, with room left for the conversation memory.
Pays off only when kept busy — avoid it for bursty development work.
Runs Olala v1.0 mini 7A1B at full precision — its 14GB of weights fit this card's 192GB, with room left for the conversation memory.
Frontier training money. If your job fits on anything cheaper, a B200 is budget burned for a spec sheet.
Runs Olala v1.0 mini 7A1B at full precision — its 14GB of weights fit this card's 96GB, with room left for the conversation memory.
Its processor is ARM, not the usual x86 — some software needs changes to run on it.
The cheapest verified way to run it
RTX 4090 at RunPod Community — $0.34/hr — runs it well, at full precision.
Launch RunPod Community ↗ See the RTX 4090 →
Checked Nov 21, 2026
Select a GPU See these GPUs on the prices board, ranked for this model.