Nemotron 3 Ultra vs Inkling
Pick a route and service tier under each model. Prices, the calculator and the verdicts follow your choice.
At a glance
Generated from the rows below
Choose Nemotron 3 Ultra for
- The lowest cost per typical request ($0.0021)
- The longest output: 131,072 tokens
- The most routes on Phaseo: 8
Choose Inkling for
- The only one that takes images, audio
The models are statistically tied on the Epoch Capabilities Index: their 95% confidence intervals overlap.
Overview
Pricing
Prices on the route and tier chosen under each model
What will it cost?
Estimate a month of requests on the route and tier chosen for each model
Uses the route and tier chosen under each model, over 30 days. Cached input is billed at the cache-read rate.
Performance
Live gateway traffic
Activity
Tokens sent through Phaseo each day over the last 30 days
None of these models has had traffic through Phaseo in the last 30 days.
Features
Features from the parameters Phaseo routes support
Benchmarks
Only results that two or more of the selected models report get a row
Epoch Capabilities Index Independent
Higher is better · dot is the estimate, band is the 95% confidence interval · Epoch AI, CC BY 4.0
Every interval overlaps, so these models are statistically tied on this index.
Artificial Analysis Intelligence Index Independent
Higher is better · at Extra high effort · 1 of 2 models tested
Where they rank
The five leaders on each capability index, then the compared models. Each model at its best-scoring setting.
Artificial Analysis Intelligence Index42 models ranked
- #1Claude Opus 5.558.4
- #2GPT-6 Astra56.9
- #3GPT-6.1 Sol54.5
- #4Claude Sonnet 5.554.2
- #5GPT-6.1 Terra51.0
- #35Inkling12.5
- #36Nemotron 3 Ultra11.8
Artificial Analysis Coding Index171 models ranked
- #1Claude Fable 5.181.6
- #2GPT 5.6 Sol78.3
- #3Claude Opus 578.0
- #4GPT-6 Astra77.1
- #5Grok 4.676.8
- #58Inkling52.1
- #64Nemotron 3 Ultra49.3
Artificial Analysis Agentic Index112 models ranked
- #1Claude Fable 5.157.9
- #2Claude Opus 556.5
- #3Qwen 3.8 Max 090256.0
- #4Muse Spark 1.355.5
- #5Qwen 3.8 Flash Next53.6
- #44Inkling22.5
- #49Nemotron 3 Ultra20.1
Epoch Capabilities Index164 models ranked
- #1GPT-6 Astra169.1
- #2Claude Opus 5.5168.4
- #3GPT-6.1 Terra168.1
- #4GPT-6.1 Sol167.8
- #5Claude Sonnet 5.5166.7
- #66Inkling148.5
- #81Nemotron 3 Ultra146.0
Lab-reported results come from each lab’s own launch materials and may use different settings. Independent results come from third-party evaluators.
Providers
Routes Phaseo serves today, standard-tier prices per 1M tokens
| Provider | Input | Output | Cache read | Context window | Max output | Tiers |
|---|---|---|---|---|---|---|
 | $0.50 | $2.15 | $0.10 | 262,144 | – | Standard |
 | $0.50 | $2.20 | $0.10 | 262,144 | 131,072 | Standard, Flex, Priority |
 | $0.50 | $2.20 | $0.13 | 262,144 | – | Standard, Batch, Flex |
 | $0.60 | $2.40 | $0.12 | 1,000,000 | 131,072 | Standard |
 | $0.60 | $2.40 | $0.12 | 262,144 | – | Standard |
 | $0.80 | $2.60 | $0.10 | 262,144 | – | Standard |
 | $0.625 | $3.125 | $0.1875 | 256,000 | 32,768 | Standard |
 | $1.00 | $3.00 | $0.12 | 1,048,576 | – | Standard |
- CoreWeave Forge
- Input
- $0.50
- Output
- $2.15
- Context window
- 262,144
Standard
- DeepInfra
- Input
- $0.50
- Output
- $2.20
- Context window
- 262,144
Standard, Flex, Priority
- Doubleword
- Input
- $0.50
- Output
- $2.20
- Context window
- 262,144
Standard, Batch, Flex
- Baseten
- Input
- $0.60
- Output
- $2.40
- Context window
- 1,000,000
Standard
- Fireworks
- Input
- $0.60
- Output
- $2.40
- Context window
- 262,144
Standard
- GMICloud
- Input
- $0.80
- Output
- $2.60
- Context window
- 262,144
Standard
- Venice
- Input
- $0.625
- Output
- $3.125
- Context window
- 256,000
Standard
- Nebius Token Factory
- Input
- $1.00
- Output
- $3.00
- Context window
- 1,048,576
Standard
Standard-tier prices. The first row is the cheapest route; pick another under the model’s name at the top.
Also listed, not routed by Phaseo: 13 providers
Crusoe, DigitalOcean, Kilo Gateway, NanoGPT, NVIDIA, OpenRouter, Parasail, Perplexity Agent, Pioneer, Requesty, Thinking Machines, Together, Vercel AI Gateway
Shown for reference: other gateways’ listings and routes Phaseo has retired or disabled.
Details
Nemotron 3 Ultra vs Inkling: side-by-side summary
Nemotron 3 Ultra and Inkling are both available through the Phaseo API, so switching between them is a change of model id rather than a new integration.
Nemotron 3 Ultra is from Nvidia. It has a 1,000,000-token context window. On its cheapest route it costs $0.50 per 1M input tokens and $2.15 per 1M output tokens.
Inkling is from Thinking Machines Lab. It has a 1,000,000-token context window. On its cheapest route it costs $1.00 per 1M input tokens and $4.05 per 1M output tokens.
Nemotron 3 Ultra model details, providers and pricingInkling model details, providers and pricing
Nemotron 3 Ultra vs Inkling at a glance
| Attribute | Nemotron 3 Ultra | Inkling |
|---|---|---|
| Lab | Nvidia | Thinking Machines Lab |
| Input price per 1M tokens | $0.50 | $1.00 |
| Output price per 1M tokens | $2.15 | $4.05 |
| Context window | 1,000,000 tokens | 1,000,000 tokens |
| Max output | 131,072 tokens | 65,536 tokens |
| Artificial Analysis Intelligence Index | 11.8 (#36) | 12.5 (#35) |
Prices are each model's cheapest standard route on Phaseo. Benchmark ranks use each model's best-scoring evaluated setting.
Frequently asked questions
Which is cheaper, Nemotron 3 Ultra or Inkling?
Nemotron 3 Ultra is cheaper for both input and output tokens. Nemotron 3 Ultra costs $0.50 per 1M input tokens and $2.15 per 1M output tokens; Inkling costs $1.00 and $4.05, each on its cheapest standard route on Phaseo.
Which has the longer context window, Nemotron 3 Ultra or Inkling?
Both have a 1,000,000-token context window.
Which scores higher on the Artificial Analysis Intelligence Index, Nemotron 3 Ultra or Inkling?
Inkling scores higher on the Artificial Analysis Intelligence Index. Inkling scores 12.5, ranked #35 of 42. Nemotron 3 Ultra scores 11.8, ranked #36 of 42.
Can I use Nemotron 3 Ultra and Inkling through one API?
Yes. Both are served through the Phaseo API, so you can switch between them by changing the model id.