The Information Machine

NVIDIA Nemotron 3 Ultra: Hybrid SSM/MoE Architecture Launch and Benchmarks · history

Version 3

2026-06-06 08:18 UTC · 130 items

What

NVIDIA launched Nemotron 3 Ultra (550B total/55B active parameters, hybrid Mamba SSM/MoE) at Computex 2026 and shipped it June 4 [1][2], simultaneously formally launching the Nemotron Coalition — a consortium of eight AI labs including Mistral AI, Nous Research, H Company, and others — with NVIDIA offering DGX Cloud compute to member labs [6][7][9]. Independent comparisons show the model delivering code output comparable to GPT-5.5 at ~10x lower cost [15], but trailing Kimi K2.6 on raw intelligence while offering 3-6x faster inference [16]. SemiAnalysis called the coalition 'communist committee-style' [21], and a deployment observation found CPU-bottlenecked latency suggesting production infrastructure is not yet optimized [20].

Why it matters

The coalition structure introduces a specific organizational model: NVIDIA subsidizes compute (DGX Cloud) in exchange for labs building on the Nemotron platform, which could shape how open-weights frontier model development is organized more broadly. The cost-performance ratio is compelling for enterprise users, but unresolved production reliability and the coalition's governance implications are the key variables determining whether the launch converts to durable adoption.

Open questions

  • Does NVIDIA's DGX Cloud subsidy to coalition member labs create dependency relationships that could constrain their independence? [9]

  • Does Nemotron 3 Ultra meet production reliability standards? A deployment observation found GPU at ~30% utilization with one CPU core pinned at 100%, suggesting per-request latency is dominated by fixed overhead rather than GPU compute [20][17].

  • Does the ~10x cost advantage over GPT-5.5 hold across task types beyond code generation, and how does the model compare to Kimi K2.6, which reportedly leads on raw intelligence? [15][16]

  • Will benchmarks beyond the GDPval-AA metric cited in NVIDIA's keynote confirm its 'most intelligent US open-weights' claim, given that agent and code-review evaluations show more modest results? [4][19][17]

Narrative

NVIDIA announced Nemotron 3 Ultra at Jensen Huang's Computex 2026 keynote and shipped the model on June 4, 2026. It has 550B total parameters with 55B active — a mixture-of-experts ratio of roughly 10:1 — and uses a hybrid architecture that pairs a Mamba-style state-space model (SSM) with a Transformer backbone [1][2]. The SSM handles long sequences without the quadratic memory cost of full attention, making the architecture suited to extended agentic reasoning and multi-step tool use. NVIDIA positioned it as the most intelligent US open-weights model, citing Artificial Analysis's GDPval-AA benchmark in the keynote [3][4]. One observer noted the release comprises not just the model but a full reproducible reasoning stack: open weights, math SFT training data, and an evaluation framework [5].

Alongside the model, NVIDIA formally launched the Nemotron Coalition, a consortium of eight AI labs committed to collaborative open frontier model development. Named members include Mistral AI, Nous Research, H Company, Prime Intellect, Black Forest Labs, Cursor AI, LangChain, and Perplexity AI [6][7][8]. NVIDIA is offering DGX Cloud compute access to coalition member labs [9]. Mistral AI's CEO Arthur Mensch posted support for the partnership at announcement [10], and Nous Research joined as a new member on launch day [11][12]. The coalition's governance structure — specifically what obligations or dependencies DGX Cloud access creates for member labs — has not been publicly detailed.

Performance comparisons since launch present a mixed picture. Artificial Analysis benchmarks show Nemotron 3 Ultra scoring 47.7 versus Qwen3.6-27B's 45.8 with faster task completion [13][14]. A direct code-generation comparison to GPT-5.5 showed nearly identical output at roughly 10x lower cost ($0.051 vs $0.57) [15]. Nemotron 3 Ultra reportedly trails Kimi K2.6 on raw intelligence while delivering 3-6x faster inference at lower cost [16]. CodeRabbit found the model close to baseline in code review but flagged production reliability as a specific concern [17][18]. One evaluator cited an agent benchmark score of 89.9% versus Claude Opus 4.7's 93.5% [19]. A deployment observation flagged GPU at ~30% utilization with one CPU core pinned at 100%, suggesting per-request latency is dominated by a large fixed cost rather than GPU computation [20].

The coalition drew immediate ideological criticism from SemiAnalysis, which on June 5 called it 'communist committee-style' and argued that free-market competition — including using Chinese open models like Kimi for open-source work — is the correct path for American open-source AI rather than coordinated consortium development [21][22]. This framing inverts the US-China narrative: SemiAnalysis describes an American industry consortium as the non-market option and a Chinese model as the free-market one. Grok/xAI separately stated it would not adopt Nemotron 3 Ultra, consistent with xAI's approach of building and training its own models end-to-end [23].

Timeline

  • 2026-06-01: Rohan Paul announces Nemotron 3 Ultra will ship within days and summarizes its hybrid SSM/MoE architecture and long-context advantages. [2]
  • 2026-06-01: Skeptical commentary emerges: Nemotron models historically rank ~5th in open-source rankings and 550B scale is called a compute showcase. [31][32]
  • 2026-06-02: Mistral AI CEO Arthur Mensch posts support for joining the Nemotron Coalition. [10]
  • 2026-06-03: Artificial Analysis notes NVIDIA cited its GDPval-AA benchmark in Jensen Huang's Computex 2026 keynote to support performance claims. [4][3]
  • 2026-06-04: Nemotron 3 Ultra officially ships; becomes available on Nebius Token Factory and through Kilocode. [34][38][39]
  • 2026-06-04: NVIDIA formally launches the Nemotron Coalition of eight AI labs, offering DGX Cloud compute access to member labs. [6][7][9]
  • 2026-06-04: Nous Research joins the Nemotron Coalition. [11][12]
  • 2026-06-04: Glean announces same-day enterprise support for Nemotron 3 Ultra. [33][35]
  • 2026-06-04: Artificial Analysis independent benchmarks: Nemotron 3 Ultra scores 47.7 versus Qwen3.6-27B's 45.8 with faster task completion. [14][13]
  • 2026-06-04: CodeRabbit evaluation: Nemotron 3 Ultra lands close to baseline with a reliability caveat for production use. [17][28][18]
  • 2026-06-04: One evaluator cites agent benchmark score of 89.9% for Nemotron 3 Ultra versus 93.5% for Claude Opus 4.7. [19]
  • 2026-06-04: Grok/xAI states it will not adopt Nemotron 3 Ultra; xAI builds and trains its own models end-to-end. [23]
  • 2026-06-05: Side-by-side comparison shows Nemotron 3 Ultra producing near-identical code output to GPT-5.5 at roughly 10x lower cost ($0.051 vs $0.57). [15]
  • 2026-06-05: SemiAnalysis calls the Nemotron Coalition 'communist committee-style' and states it will use Chinese open models (Kimi) instead for open-source work. [21][22]
  • 2026-06-05: Infomly reports Nemotron 3 Ultra trails Kimi K2.6 on raw intelligence but delivers 3-6x faster inference at lower cost. [16]
  • 2026-06-05: Deployment observation: GPU at ~30% utilization, one CPU core pinned at 100%, per-request latency dominated by fixed overhead. [20]
  • 2026-06-05: Nemotron 3 Ultra added to HuggingChat, expanding public access beyond inference API providers. [36]

Perspectives

NVIDIA (official)

Nemotron 3 Ultra is the most intelligent US open-weights model built for long-running agentic workloads; the Nemotron Coalition advances American open-source frontier AI through lab collaboration with DGX Cloud compute support.

Evolution: Consistent with NVIDIA's Nemotron positioning; the coalition announcement adds an organizational dimension to what was previously a single-model story.

[24][25][6][9][26]

Artificial Analysis (@ArtificialAnlys)

Independent benchmarks confirm Nemotron 3 Ultra edges Qwen3.6-27B (47.7 vs 45.8) with faster task completion; their GDPval-AA metric was cited in NVIDIA's keynote.

Evolution: Consistent empirical testing role; elevated to keynote prominence by NVIDIA's citation.

[14][4][3][27]

Rohan Paul (@rohanpaul_ai)

Positive on architectural rationale and cost case: SSM enables long-context processing without attention bottlenecks, and the ~10x cost gap versus GPT-5.5 on code tasks is compelling for developers.

Evolution: Consistent; provided pre-release architecture summary and followed up with a direct GPT-5.5 price comparison on launch day.

[2][15]

CodeRabbit (@coderabbitai)

Measured: Nemotron 3 Ultra performs close to baseline in code review benchmarks but production reliability is a specific concern for production workloads.

Evolution: Consistent; full blog post confirms the reliability caveat raised at launch.

[17][28][18][29]

SemiAnalysis (@SemiAnalysis_)

The Nemotron Coalition is 'communist committee-style' and the wrong model for American open-source AI; free-market competition including use of Chinese open models like Kimi is preferable.

Evolution: Consistent since June 5 entry into thread; coalition membership details in new items reinforce rather than change their critique.

[21][22]

Mistral AI and coalition members

Supportive of open frontier model collaboration with NVIDIA; Mistral's CEO expressed enthusiasm for building frontier open-source AI together with NVIDIA as a coalition member.

Evolution: New voice; coalition membership formalized and publicly endorsed at launch.

[30][10][11][12]

Skeptics (ECLresearch, Aero, Infomly)

Nemotron 3 Ultra is primarily a speed and cost showcase; Nemotron models historically rank ~5th in open-source, and Kimi K2.6 leads on raw intelligence even as Nemotron is faster and cheaper.

Evolution: Skeptical position updated by Infomly's Kimi K2.6 comparison, adding a specific capability counter-reference that wasn't present at launch.

[31][32][16]

Enterprise integrators (Glean, Nebius, Simplismart)

Enthusiastic same-day and near-day adoption; framing Nemotron 3 Ultra as expanding enterprise model choice for agentic applications.

Evolution: Consistent; availability expanded to HuggingChat and Simplismart since launch day.

[33][34][35][36][37]

Tensions

  • NVIDIA claims Nemotron 3 Ultra is the most intelligent US open-weights model; skeptics argue Nemotron historically ranks ~5th in open-source and Kimi K2.6 leads on raw intelligence, even as Nemotron offers faster inference at lower cost. [24][31][16]
  • NVIDIA's keynote cited GDPval-AA to support performance claims; CodeRabbit's code review benchmark and an agent benchmark score (89.9% vs Claude Opus 4.7's 93.5%) show more modest results than the headline framing suggests. [4][17][19]
  • A code-generation comparison shows ~10x cost savings versus GPT-5.5, but a deployment observation found CPU-bottlenecked latency and CodeRabbit flagged production reliability, qualifying the cost-advantage case for production use. [15][20][17]
  • SemiAnalysis argues the Nemotron Coalition's coordinated structure contradicts American free-market principles; NVIDIA frames the coalition as advancing American open-source AI through DGX Cloud support and lab collaboration. [21][22][6][9]
  • SemiAnalysis prefers Chinese open models (Kimi) over an American-organized coalition, inverting the US-China framing by labeling the consortium as the non-market option; Infomly separately confirms Kimi K2.6 does lead Nemotron on raw intelligence. [22][16]

Sources

  1. [1] NVIDIA has announced Nemotron 3 Ultra at Jensen Huang's Computex 2026 keynote. At 550 billion total parameters (55 billi... — reactive:nvidia-nemotron-ultra (2026-06-04)
  2. [2] Nemotron 3 Ultra will be available from Nvidia in few days. — Rohan Paul Twitter (2026-06-01)
  3. [3] Jensen Huang’s keynote at Computex used Artificial Analysis benchmarks to communicate the performance of Nemotron 3 Ultr... — reactive:nvidia-nemotron-ultra (2026-06-03)
  4. [4] NVIDIA used the GDPval-AA benchmark to promote the launch of NVIDIA Nemotron 3 Ultra https://t.co/N6e4ItzYev — reactive:nvidia-nemotron-ultra (2026-06-03)
  5. [5] NVIDIA is not merely releasing Nemotron 3 Ultra. It is releasing a reproducible reasoning stack: open weights, math SFT ... — reactive:nvidia-nemotron-ultra (2026-06-05)
  6. [6] NVIDIA Launches Nemotron Coalition of Leading Global AI Labs to Advance Open Frontier Models | NVIDIA Newsroom — reactive:open-model-capability-gap
  7. [7] Nvidia's Nemotron coalition brings eight AI labs together to build open frontier models | Tom's Hardware — reactive:open-model-capability-gap
  8. [8] @NVIDIAAI @hcompany_ai @NousResearch @PrimeIntellect @bfl_ai @cursor_ai @LangChain @MistralAI @perplexity_ai @Reflection... — reactive:nvidia-nemotron-ultra (2026-06-05)
  9. [9] Nvidia offers DGX Cloud to AI foundation model labs for new ... — reactive:nvidia-nemotron-ultra
  10. [10] RT @arthurmensch: Looking forward to building frontier open source AI models together with @Nvidia as we join the Nemotr... — reactive:nvidia-nemotron-ultra (2026-06-02)
  11. [11] Nous Research (AI lab open-source) baru join coalition Nvidia buat model Nemotron. — reactive:nvidia-nemotron-ultra (2026-06-05)
  12. [12] Excited to see @NousResearch join NVIDIA's Nemotron Coalition! — reactive:nvidia-nemotron-ultra (2026-06-04)
  13. [13] Not only did Nemotron 3 Ultra beat Qwen3.6-27b in raw AA benchmark (47.7 vs 45.8) but it also completed the tasks faster... — reactive:nvidia-nemotron-ultra (2026-06-04)
  14. [14] Nemotron 3 Ultra was launched today, including a focus on low latency agentic performance. We tested it against peers un... — reactive:nvidia-nemotron-ultra (2026-06-04)
  15. [15] Nemotron 3 Ultra vs GPT-5.5 on atomic[.]chat, a desktop app that runs LLMs locally. — Rohan Paul Twitter (2026-06-05)
  16. [16] NVIDIA Nemotron 3 Ultra Trails Kimi K2.6 on raw intelligence but delivers 3-6x faster inference at lower cost. — reactive:nvidia-nemotron-ultra (2026-06-05)
  17. [17] In our CodeRabbit benchmark, Nemotron 3 Ultra landed close to the baseline: — reactive:nvidia-nemotron-ultra (2026-06-04)
  18. [18] Nemotron 3 Ultra makes the case for fast, open coding models — reactive:nvidia-nemotron-ultra
  19. [19] @turingou 我一直堅信固定工作流的agent不需要SoTA，英偉達nemotron 3 ultra在agent benchmark 拿了89.9%，OPUS 4.7拿了93.5%。 — reactive:nvidia-nemotron-ultra (2026-06-04)
  20. [20] GPU at ~30% utilization. One CPU core pinned at 100%. Per-request latency dominated by a large fixed cost. — reactive:nvidia-nemotron-ultra (2026-06-05)
  21. [21] The path for American open models to catch up to Chinese open models is not through a “coalition,” but through the found… — SemiAnalysis Twitter (2026-06-05)
  22. [22] We fundamentally disagree with the communist committee-style “Nemotron Coalition” approach to developing OSS models, and… — SemiAnalysis Twitter (2026-06-05)
  23. [23] @xiaosun86 @NVIDIAAI No, Grok isn't adopting Nemotron 3 Ultra. xAI builds and trains its own models end-to-end. Strong b... — reactive:nvidia-nemotron-ultra (2026-06-01)
  24. [24] Nemotron 3 Ultra 550B (A55B) is the largest Nemotron 3 model to date, and the most intelligent US open weights model rig... — reactive:nvidia-nemotron-ultra (2026-06-01)
  25. [25] NVIDIA ships Nemotron 3 Ultra, a 550B MoE model for long-running agents. It runs 5x faster at inference and reduces cost... — reactive:nvidia-nemotron-ultra (2026-06-04)
  26. [26] NVIDIA Nemotron 3 Ultra Powers Faster, More Efficient Reasoning ... — reactive:nvidia-nemotron-ultra
  27. [27] Nemotron 3 Ultra - Intelligence, Performance & Price Analysis — reactive:nvidia-nemotron-ultra
  28. [28] The caveat: reliability. — reactive:nvidia-nemotron-ultra (2026-06-04)
  29. [29] Nemotron 3 Ultra makes the case for fast, open coding models — reactive:nvidia-nemotron-ultra
  30. [30] Mistral AI partners with NVIDIA to accelerate open frontier models — reactive:open-model-capability-gap
  31. [31] @jun_song Solid skepticism. Nemotron 3 Ultra at 550B is a compute flex, but the open-weight ecosystem moved past "just b... — reactive:nvidia-nemotron-ultra (2026-06-01)
  32. [32] @scaling01 Nemotron models are always released to be #5 on open-source while using the latest nvidia tricks to be fast a... — reactive:nvidia-nemotron-ultra (2026-06-01)
  33. [33] 🎉 We’re announcing support for @nvidia Nemotron Ultra 3! — reactive:nvidia-nemotron-ultra (2026-06-04)
  34. [34] NVIDIA Nemotron™ 3 Ultra is now live on Nebius Token Factory. — reactive:nvidia-nemotron-ultra (2026-06-04)
  35. [35] Excited to expand model choice in @Glean with @NVIDIA Nemotron 3 Ultra. — reactive:nvidia-nemotron-ultra (2026-06-04)
  36. [36] Nemotron 3 Ultra is on HuggingChat ☘️ — reactive:nvidia-nemotron-ultra (2026-06-05)
  37. [37] NVIDIA Nemotron 3 Ultra Now Available on Simplismart: Tailored Inference for Agentic AI — reactive:nvidia-nemotron-ultra
  38. [38] Nemotron 3 ultra, Qwen 3.7 plus, laguna m.1, step 3.7 flash ALL FREE on kilocode. but which to use? BENCHMARK TEST TIME.... — reactive:nvidia-nemotron-ultra (2026-06-04)
  39. [39] 550 billion parameters. Open weights. Ships today. — reactive:nvidia-nemotron-ultra (2026-06-04)