Size
187.8B
a dense model — all of it works on every answer
Agnes 2.5 Pro Theta
Context
262K
the longest conversation the model is built to hold
Lab
Agnes AI — Agnes
the lab that released it
Released
Nov 15, 2026
Licence: Apache 2.0 — free for any use, including commercial. Just keep the licence notice.
Worth knowing
Agnes AI's multimodal MoE, weights published on Hugging Face on Nov 15, 2026 under Apache 2.0 after an API-only period: 187.8B total parameters, 512 routed experts with top-10 routing, configured 262K context, BF16 payload 375.5 GB — the card names 4 × H200-class GPUs as the practical starting point. Publisher-reported benchmark snapshot (not a reproducible bundle, per the card itself): GPQA Diamond 92.9, LiveCodeBench v6 90.5. Custom Transformers code (trust_remote_code=True); optimised serving-engine compatibility not yet established.
What it needs
Full precision (FP16)
376 GB
the weights, uncompressed
8-bit
188 GB
the weights, compressed to 8-bit
4-bit
94 GB
the weights, compressed to 4-bit
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 Agnes 2.5 Pro Theta compressed — the 8-bit build needs 188GB, which fits this card's 288GB. The compromise: slightly lower answer quality and a shorter conversation memory than the full-precision build.
Very costly for small jobs — best value only on big work.
Runs Agnes 2.5 Pro Theta compressed — the 4-bit build needs 94GB, which fits this card's 141GB. The compromise: slightly lower answer quality and a shorter conversation memory than the full-precision build.
More card than most jobs need; the main trap is paying for it on small work.
Runs Agnes 2.5 Pro Theta compressed — the 4-bit build needs 94GB, which fits this card's 192GB. The compromise: slightly lower answer quality and a shorter conversation memory than the full-precision build.
Frontier training money. If your job fits on anything cheaper, a B200 is budget burned for a spec sheet.
At full precision, Agnes 2.5 Pro Theta needs more than one GPU — no single card holds its 376GB of weights plus room to work. Every card above runs a compressed build.
The cheapest verified way to run it
H200 at RunPod Community — $3.59/hr — runs it compressed — the 4-bit build.
Launch RunPod Community ↗ See the H200 →
Checked Nov 15, 2026
Select a GPU See these GPUs on the prices board, ranked for this model.