MDL
Llama 3.1 Nemotron Ultra 253B v1
NVIDIA
Verified against Artificial Analysis · Jul 13, 2026
Capability
56
percentile index
Price /M
$0.6 / $1.8
in / out
Context
—
Avg score
43.7
11 benchmarks
AA Intelligence
9.1
index
Speed
52 tok/s
TTFT 0.7s
Released
Apr 7, 2025
About
Llama 3.1 Nemotron Ultra 253B v1 is a open-weights model in the Llama family from NVIDIA. Benchmarked on 11 evals, averaging 43.7.
Benchmark scores · 11
avg 43.7 — higher is betterThe research behind this model
all papers mentioning it →Information-Aware KV Cache Compression for Long Reasoning▲ 11 on HF · Jun 25, 2026RoPE-Aware Bit Allocation for KV-Cache Quantization▲ 8 on HF · Jun 23, 2026Off-the-Shelf LLMs as Process Scorers: Training-Free Alternative to PRMs for Mathematical Reasoning▲ 7 on HF · Jun 1, 2026Self-Guided Test-Time Training for Long-Context LLMs▲ 2 on HF · Jul 10, 2026When Agents Commit Too Soon: Diagnosing Premature Commitment in LLM Agents▲ 2 on HF · Jun 22, 2026
Mentions matched by name in title/abstract — from the arXiv + HF daily corpus.
Performance & price
Output speed52 tok/s
Latency (TTFT)0.7s
First answer39s
Price in / out$0.6 / $1.8 per 1M
Blended price$0.9 / 1M
Math index63.7
Median across API providers, via Artificial Analysis. Blended = 3:1 input:output. For reasoning models, latency includes thinking time. Methodology →
History
full ledger →Capability over time
Append-only ledger — every observed change to this model's numbers.
API providers
Artificial Analysis$1.8/M