---
format: "aidr-story-markdown/v1"
id: "4773a8bbcebba4b029b7cc632a2360b16ec1bc710d431107694c38f0283ec66c"
canonical_url: "https://aidr.today/4773a8bb?lang=en"
title: "AI21 Beats Claude Opus 5.5 at 33x Lower Cost with Nemotron 3.5"
lang: "en"
requested_lang: "en"
available_langs: ["en","vi"]
translation_fallback: null
fallback_fields: []
published_at: "2026-10-15T15:39:45.000Z"
category: "Models"
topics: ["nvidia","llm"]
source_urls: ["https://marketbrief.now/ai/ai21-beats-claude-opus-55-at-33x-lower-cost-with-nemotron-35-157dc711","https://huggingnews.com/ai/ai21-beats-claude-opus-55-at-33x-lower-cost-with-nemotron-35-157dc711"]
summary: "AI21 Labs collaborated with firms like Gong and Deel to post-train NVIDIA's Nemotron 3.5 Super VL model into specialized corporate tools, achieving performance comparable to general frontier AI at lower cost. A Deel-specific model reached 88.5% accuracy on an 8-country payroll benchmark, outperforming Claude Opus 5.5 by 8% while running at 33x lower cost."
---

# AI21 Beats Claude Opus 5\.5 at 33x Lower Cost with Nemotron 3\.5

> [Open the canonical story](<https://aidr.today/4773a8bb?lang=en>)

**Published:** 2026-10-15T15:39:45.000Z
**Category:** Models
**Topics:** nvidia, llm

## Summary

AI21 Labs collaborated with firms like Gong and Deel to post\-train NVIDIA's Nemotron 3\.5 Super VL model into specialized corporate tools, achieving performance comparable to general frontier AI at lower cost\. A Deel\-specific model reached 88\.5% accuracy on an 8\-country payroll benchmark, outperforming Claude Opus 5\.5 by 8% while running at 33x lower cost\.

## Sources

- [Story source](<https://marketbrief.now/ai/ai21-beats-claude-opus-55-at-33x-lower-cost-with-nemotron-35-157dc711>)
- [Story source](<https://huggingnews.com/ai/ai21-beats-claude-opus-55-at-33x-lower-cost-with-nemotron-35-157dc711>)