Agnes 2.5 Pro Theta vs Nemotron 3 Super 120B A12B
Full profiles: Agnes 2.5 Pro Theta (Sapiens AI) and Nemotron 3 Super 120B A12B (NVIDIA): every benchmark, price and rank.
Updated · Data from 3 sources
As of 16 Nov 2026, Agnes 2.5 Pro Theta and Nemotron 3 Super 120B A12B are neck and neck on quality (50.8 vs 51.1 on the BenchLeader Index, inside the ±14.4 margin), and Agnes 2.5 Pro Theta is 3.0× cheaper at $0.150 vs $0.450 per 1M tokens blended. Agnes 2.5 Pro Theta leads in every category we measure both in. Agnes 2.5 Pro Theta streams faster.
Key facts (as of 16 Nov 2026)
- Agnes 2.5 Pro Theta scores 50.8 on the BenchLeader Index and Nemotron 3 Super 120B A12B scores 51.1 (BenchLeader, 16 Nov 2026).
- Agnes 2.5 Pro Theta costs $0.150 per 1M tokens blended (list price, 16 Nov 2026).
- Nemotron 3 Super 120B A12B costs $0.450 per 1M tokens blended (list price, 16 Nov 2026).
- Agnes 2.5 Pro Theta streams faster than Nemotron 3 Super 120B A12B (Artificial Analysis, 7-day median, 16 Nov 2026).
- Of the 9 benchmarks both models have, Agnes 2.5 Pro Theta leads 9 (BenchLeader, 16 Nov 2026).
Which to pick for…
The pick's numbers come first. Category scores on the BenchLeader scale (50 is the average model, 15 points one standard deviation). High confidence means a lead of half a standard deviation or more; under one point is called either.
| Use case | Pick | Why | Confidence |
|---|---|---|---|
| Reasoning | Agnes 2.5 Pro Theta | 44 vs 35 category score; HLE-Diamond (AA) 16.8% vs 4.9% | High |
| Long documents | Agnes 2.5 Pro Theta | 67 vs 57 category score; AA-LCR v1.1 84.7% vs 65.7% | High |
| Lowest cost | Agnes 2.5 Pro Theta | $0.150 vs $0.450 per 1M blended; agentic coding session $0.007 vs $0.022 | – |
| Fastest first answer | Agnes 2.5 Pro Theta | 8.1 s (Agnes 2.5 Pro Theta) vs 10.7 s (Nemotron 3 Super 120B A12B (thinking)) to the first answer token | – |
| High volume on a budget | Neither flagship | A smaller model is usually the better fit: Cheapest good LLMs, Fastest LLMs | – |
Key differences
- Agnes 2.5 Pro Theta and Nemotron 3 Super 120B A12B (thinking) are neck and neck on quality (50.8 vs 51.1, within the ±14.4 margin).
- Agnes 2.5 Pro Theta is stronger in long context, reasoning.
- Agnes 2.5 Pro Theta is 3.0× cheaper ($0.150 vs $0.450 per 1M blended).
At a glance
| Spec | Agnes 2.5 Pro Theta | Nemotron 3 Super 120B A12B |
|---|---|---|
| Developer | Sapiens AI | NVIDIA |
| Released | 7 Nov 2026 | 11 Mar 2026 |
| Context window (tokens) | 1,000,000 | 262,144 |
| Compared at | Agnes 2.5 Pro Theta | Nemotron 3 Super 120B A12B (thinking) |
| Open weights | No | Yes |
| Price in / out / cached input per 1M | $0.100 / $0.300 / – | $0.300 / $0.900 / $0.060 |
Category by category
Category scores on the BenchLeader scale: 50 is the average evaluated model, 15 points is one standard deviation. A gap under one point is called level.
| Category | Agnes 2.5 Pro Theta | Nemotron 3 Super 120B A12B (thinking) | Leader |
|---|---|---|---|
| Reasoning | 43.7 | 34.7 | Agnes 2.5 Pro Theta |
| Coding | 45.7 | – | not comparable |
| Agents & tools | – | 44.9 | not comparable |
| Knowledge | 38.7 | – | not comparable |
| Instruction following | – | 70.2 | not comparable |
| Multimodal | 60.8 | – | not comparable |
| Long context | 67.1 | 57.2 | Agnes 2.5 Pro Theta |
On one chart
Each domain on the index's own scale, where 50 is an average model. Click a domain to see the benchmarks behind it. Pick up to 5 models from the list.
Dashed: that model has no result on the axis it crosses; the line only joins its neighbours.
Not drawn, no result for these models: Maths, Human preference.
Where sources disagree
Across the 9 benchmarks both have, Agnes 2.5 Pro Theta leads 9, Nemotron 3 Super 120B A12B leads 0.
| Overall index | Agnes 2.5 Pro Theta | Nemotron 3 Super 120B A12B (thinking) | Leader |
|---|---|---|---|
| BenchLeader Index | 50.8 | 51.1 | neck and neck |
Each publisher runs its own tasks, harnesses and effort settings, so their overall verdicts can differ; BenchLeader reconciles them on one scale. † measured at another effort setting of the same model (the publisher did not test the one compared here).
Price, speed and latency
List prices blended at three input tokens per output token. Per-task prices use the token counts on the methodology page. Cost, tokens and time per benchmark task are measured: Artificial Analysis’ averages over its whole Intelligence Index run.
| Measure | Agnes 2.5 Pro Theta | Nemotron 3 Super 120B A12B (thinking) |
|---|---|---|
| Blended price per 1M tokens | $0.150 | $0.450 |
| Input / output per 1M | $0.100 / $0.300 | $0.300 / $0.900 |
| Cached input per 1M | – | $0.060 |
| Price per task: chat reply | $0.0001 | $0.0004 |
| Price per task: summarize a 30-page report | $0.0014 | $0.0041 |
| Price per task: code edit | $0.0011 | $0.0032 |
| Price per task: agentic coding session | $0.0072 | $0.022 |
| Price per task: structured extraction | $0.0003 | $0.0008 |
| Agentic coding session, input mostly cached | – | $0.011 |
| Cost per benchmark task (AA Index) | $0.575 | – |
| Output tokens per task (AA Index) | 278k | – |
| Of which reasoning | 229k | – |
| Time per task (AA Index) | 24 min | – |
| Output speed | 190 tok/s | 183 tok/s |
| Time to first answer | 8.1 s | 10.7 s |
| Time to a complete answer, last measured 13 Nov 2026 | 16.9 s | 38.6 s |
Each model is compared at its best-ranked reasoning-effort setting. Agnes 2.5 Pro Theta scores 50.8 and takes 8 s to its first answer token. Nemotron 3 Super 120B A12B (thinking) scores 51.1 and takes 11 s to its first answer token. Higher effort means more thinking before the answer: better scores, longer waits and more tokens billed. How reasoning effort changes scores and cost.
Every benchmark
39 benchmarks, grouped by category; each name links to the board and its leaders. † measured at another effort setting of the same model, shown for reference and not counted as a win.
| Metric | Publisher | Agnes 2.5 Pro Theta | Nemotron 3 Super 120B A12B (thinking) |
|---|---|---|---|
| BenchLeader Index | BenchLeader | 50.8 | 51.1 |
| Coding score | BenchLeader | 45.7 | – |
| Knowledge score | BenchLeader | 38.7 | – |
| Long context score | BenchLeader | 67.1 | 57.2 |
| Multimodal score | BenchLeader | 60.8 | – |
| Reasoning score | BenchLeader | 43.7 | 34.7 |
| Agents & tools score | BenchLeader | – | 44.9 |
| Instruction following score | BenchLeader | – | 70.2 |
| Blended price $/M | BenchLeader | $0.150 | $0.450 |
| Output speed | BenchLeader | 190 tok/s | 183 tok/s |
| Time to first answer | BenchLeader | 8.1 s | 10.7 s |
| Context window | BenchLeader | 1,000,000 tokens | 262,144 tokens |
| Reasoning | |||
| Arena Hard Prompts | Arena (LMArena) | – | 1378† |
| GPQA Diamond (AA) | Artificial Analysis | – | 80.0% |
| Humanity's Last Exam (AA) | Artificial Analysis | 37.2% | 20.8% |
| HLE-Diamond (AA) | Artificial Analysis | 16.8% | 4.9% |
| CritPt | Artificial Analysis | 15.1% | 3.1% |
| CritPt v2 hard subset (AA) | Artificial Analysis | 26.0% | – |
| Arena Expert | Arena (LMArena) | – | 1394† |
| Coding | |||
| Arena Coding | Arena (LMArena) | – | 1407† |
| SciCode (AA) | Artificial Analysis | 56.0% | 36.2% |
| Terminal-Bench 4.0 (AA) | Artificial Analysis | 10.1% | 0.0% |
| Terminal-Bench 2.1 (AA) | Artificial Analysis | – | 38.6% |
| FrontierCode 2.0 (AA) | Artificial Analysis | 16.6% | – |
| Agents & tools | |||
| Terminal-Bench Hard | Artificial Analysis | – | 28.8% |
| τ²-Bench Telecom (AA) | Artificial Analysis | – | 67.8% |
| GDPval-AA v2.1 | Artificial Analysis | 1507 | 497 |
| τ³-Banking (AA) | Artificial Analysis | – | 10.3% |
| ITBench SRE (AA) | Artificial Analysis | – | 1.1% |
| APEX-Agents (AA) | Artificial Analysis | – | 1.8% |
| AutomationBench | Artificial Analysis | 48.8% | 3.8% |
| GDP.pdf (AA) | Artificial Analysis | 14.2% | 2.6% |
| EnterpriseOps-Gym | Artificial Analysis | – | 23.6% |
| AA-Briefcase v1.1 | Artificial Analysis | 1496 | – |
| Terminal-Bench-Science 0.1 (AA) | Artificial Analysis | 3.8% | – |
| Maths | |||
| Arena Maths | Arena (LMArena) | – | 1370† |
| Knowledge | |||
| AA-Omniscience v1.1 | Artificial Analysis | -11.2 | – |
Show all 39 benchmarks
| Metric | Publisher | Agnes 2.5 Pro Theta | Nemotron 3 Super 120B A12B (thinking) |
|---|---|---|---|
| Knowledge | |||
| AA-Omniscience v1.1: accuracy | Artificial Analysis | 21.3% | – |
| AA-Omniscience v1.1: non-hallucination | Artificial Analysis | 58.6% | – |
| Instruction following | |||
| IFBench | Artificial Analysis | – | 71.5% |
| Arena Instruction Following | Arena (LMArena) | – | 1342† |
| Human preference | |||
| Arena Text | Arena (LMArena) | – | 1360† |
| Arena Creative Writing | Arena (LMArena) | – | 1302† |
| Arena Multi-turn | Arena (LMArena) | – | 1349† |
| Arena Longer Queries | Arena (LMArena) | – | 1357† |
| Multimodal | |||
| MMMU-Pro | Artificial Analysis | 78.4% | – |
| Long context | |||
| AA-LCR v1.1 | Artificial Analysis | 84.7% | 65.7% |
| Composite | |||
| AA Intelligence Index v5.0.1 | Artificial Analysis | 24.0 | – |
| AA Intelligence Index v4.3.2 | Artificial Analysis | – | 12.8 |
| AA Progress Index v1 | Artificial Analysis | – | 233.7 |
| Metadata | |||
| AA Openness Index | Artificial Analysis | – | 83.3 |
Data as of 2026-11-16. Best configuration of each model; every score links to its source on the model pages.
What changed
- Nemotron 3 Super 120B A12B (thinking): 1 new result (GDPval-AA v2.1).
- Agnes 2.5 Pro Theta: first results.
- Nemotron 3 Super 120B A12B (thinking): BenchLeader Index 52.4 → 51.1.
- Nemotron 3 Super 120B A12B (thinking): 2 new results (AA Progress Index v1, HLE-Diamond (AA)).
- Nemotron 3 Super 120B A12B (thinking): cached input price moved from $0.30/M to $0.060/M (80% down).
- Nemotron 3 Super 120B A12B (thinking): BenchLeader Index 52.9 → 52.4.
- Nemotron 3 Super 120B A12B (thinking): 4 new results (AutomationBench, GDP.pdf (AA), MLCR…).
Related comparisons
- Agnes 2.5 Pro Theta vs GLM 4.7 Flash
- Agnes 2.5 Pro Theta vs Ling 3.1 Flash
- Agnes 2.5 Pro Theta vs Qwen3.5-9B
Rankings: Best LLM overall · Best LLM for agents · Best cheap LLMs
Agnes 2.5 Pro Theta vs Nemotron 3 Super 120B A12B: questions
- Is Agnes 2.5 Pro Theta better than Nemotron 3 Super 120B A12B?
- Agnes 2.5 Pro Theta and Nemotron 3 Super 120B A12B (thinking) are neck and neck on quality (50.8 vs 51.1, within the ±14.4 margin). The BenchLeader Index combines every independent quality benchmark; the two are within the margin of error as of 16 Nov 2026, so check the category scores for your use.
- Agnes 2.5 Pro Theta vs Nemotron 3 Super 120B A12B: which is cheaper, and will that change?
- Agnes 2.5 Pro Theta is cheaper today: $0.150 against $0.450 per million tokens, blended at three input tokens per output token. No price change has been announced for either.
- Which is faster, Agnes 2.5 Pro Theta or Nemotron 3 Super 120B A12B?
- Agnes 2.5 Pro Theta streams faster (190 vs 183 output tokens per second, Artificial Analysis 7-day median).
- Which is better for long documents?
- Agnes 2.5 Pro Theta scores higher on long-context benchmarks (67 vs 57). Agnes 2.5 Pro Theta accepts more context: 1,000,000 against 262,144 tokens for Nemotron 3 Super 120B A12B.
- Which has lower latency?
- Agnes 2.5 Pro Theta answers first: 8.1 s against 10.7 s to the first answer token, as Artificial Analysis measures it. Reasoning effort drives most of this: a higher setting thinks longer before it answers.
More questions (2)
- Do independent benchmarks agree on Agnes 2.5 Pro Theta vs Nemotron 3 Super 120B A12B?
- Across the 9 benchmarks both have, Agnes 2.5 Pro Theta leads 9, Nemotron 3 Super 120B A12B leads 0.
- When was Agnes 2.5 Pro Theta released?
- Agnes 2.5 Pro Theta was released on 7 Nov 2026.