SambaNova: Models Intelligence, Performance & Price
This analysis is intended to support you in choosing the best model provided by SambaNova for your use-case.
Most Intelligent
MiniMax-M3.1-FlashIntelligence Index
Total 7 models
Fastest
gpt-oss-120b (low)gpt-oss-120b (high)
Gemma 4 31BGemma 4 31B (non-reasoning)MiniMax-M3.1-FlashOutput speed
Total 7 models
Lowest Price
MiniMax-M3.1-Flashgpt-oss-120b (low)gpt-oss-120b (high)Gemma 4 31BGemma 4 31B (non-reasoning)Blended price (per 1M tokens)
Total 7 models
SambaNova offers 7 models, each with different intelligence, performance, and pricing characteristics. Below is a comparison of the key metrics across models.
- For output speed, the fastest models are gpt-oss-120b (low) (740 t/s), gpt-oss-120b (high) (707 t/s), and Gemma 4 31B (477 t/s). Speed varies significantly across models, with a 289% difference between the fastest and slowest.
- For latency, Gemma 4 31B (non-reasoning) (3.04s), gpt-oss-120b (low) (3.89s), and gpt-oss-120b (high) (3.99s) offer the lowest time to first answer token.
- For pricing, MiniMax-M3.1-Flash ($0.20), gpt-oss-120b (low) ($0.26), and gpt-oss-120b (high) ($0.26) offer the lowest blended prices per 1M tokens. Prices vary up to 2.3x across models.
- For context window size, MiniMax-M3.1-Flash (1M), gpt-oss-120b (low) (131k), and Gemma 4 31B (131k) support the largest context windows on SambaNova.
Intelligence Evaluations
Artificial Analysis Intelligence Index
Evaluations
Agentic knowledge work, Elo
Knowledge
1 - hallucination rate
Agentic SaaS workflows
Agentic scientific research workflows in a terminal
Agentic software engineering in open-source repositories
Reasoning & knowledge
Professional document reasoning, All-pass
Agentic coding & terminal use
Physics reasoning
Quantitative analysis on spreadsheets & documents
Legal agentic work, Hallucination-Gated All-Pass Rate
Agentic business operations
Agentic real-world work tasks, Elo
Agentic interactive 3D experiences, Elo
Product and brand videos designed in code, Elo
Kubernetes incident root-cause analysis
Medical long context reasoning
Visual reasoning
Rich, varied, well-crafted language
Well-structured, scannable, answer-first
Truthful, acknowledges uncertainty, non-sycophantic
Use of filler and generic language
Intelligence Index vs. Price
Context Window
Context Window
Pricing
Intelligence Index vs. Price
Performance Summary
Output Speed vs. Price
Speed
Measured by Output Speed (tokens per second)
Output Speed
Latency
Measured by Time (seconds) to First Token
Latency: Time To First Answer Token
Cache Behaviour
Cache Hit Rate
Cost per Task vs. Cache Hit Rate
End-to-End Response Time
Seconds to output 500 tokens, calculated based on time to first token, 'thinking' time for reasoning models, and output speed
End-to-End Response Time vs. Price
Further Analysis | |||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
MiniMax-M3.1-Flash | 1.05M | Open | 29 | $2.97 | 190 | 2.14 | 15.29 | 10.52 | |||
 DeepSeek V3.1 Terminus | 131k | Open | -- | -- | -- | -- | -- | -- | |||
DeepSeek V3.1 | 128k | Open | -- | -- | 118 | 4.31 | 25.54 | 16.99 | |||
DeepSeek R1 Distill Llama 70B | 131k | Open | -- | -- | -- | -- | -- | -- | |||
DeepSeek V3.2 (non-reasoning) | 8.19k | Open | -- | -- | -- | -- | -- | -- | |||
 Llama 3.3 70B | 128k | Open | -- | -- | -- | -- | -- | -- | |||
 Qwen3 32B | 32.8k | Open | -- | -- | -- | -- | -- | -- | |||
Qwen3 32B (non-reasoning) | 32.8k | Open | -- | -- | -- | -- | -- | -- | |||
gpt-oss-120b (low) | 131k | Open | -- | -- | 740 | 1.19 | 4.57 | 2.70 | |||
Gemma 4 31B | 131k | Open | -- | -- | 477 | 8.40 | 13.09 | 3.64 | |||
Gemma 4 31B (non-reasoning) | 131k | Open | -- | -- | 202 | 3.04 | 5.52 | -- | |||
DeepSeek V3.2 | 8.19k | Open | -- | -- | -- | -- | -- | -- | |||
DeepSeek V3.1 Terminus (non-reasoning) | 128k | Open | -- | -- | -- | -- | -- | -- | |||
gpt-oss-120b (high) | 131k | Open | -- | -- | 707 | 1.16 | 4.69 | 2.83 | |||
DeepSeek V3.1 (non-reasoning) | 131k | Open | -- | -- | 185 | 5.00 | 7.71 | -- | |||
Key definitions
Frequently Asked Questions
Common questions about SambaNova
SambaNova offers 7 models that we track: MiniMax-M3.1-Flash, DeepSeek V3.1, gpt-oss-120b (low), Gemma 4 31B, Gemma 4 31B (non-reasoning), gpt-oss-120b (high), and DeepSeek V3.1 (non-reasoning).
The most intelligent model available on SambaNova is MiniMax-M3.1-Flash with an Intelligence Index score of 29.
The fastest model on SambaNova by output speed is gpt-oss-120b (low) at 739.9 tokens per second.
The model with the lowest time to first answer token on SambaNova is Gemma 4 31B (non-reasoning) at 3.04s. Lower latency means faster initial response time.
The most affordable model on SambaNova by blended price is MiniMax-M3.1-Flash at $0.20 per 1M tokens (7:2:1 cache hit/input/output ratio).
Prices on SambaNova vary up to 16x across models, from $0.20 per 1M tokens for MiniMax-M3.1-Flash to $3.15 per 1M tokens for DeepSeek V3.1 (non-reasoning).
Yes, SambaNova offers an OpenAI-compatible API, making it easy to switch from OpenAI or use existing OpenAI SDK integrations.
Yes, all 7 models on SambaNova support JSON mode for structured output.
Yes, all 7 models on SambaNova support function calling (tool use).
Yes, SambaNova offers 5 reasoning models: MiniMax-M3.1-Flash, DeepSeek V3.1, gpt-oss-120b (low), Gemma 4 31B, and gpt-oss-120b (high). Reasoning models use extended thinking to work through complex problems before providing an answer.
Yes, all 7 models on SambaNova are open weight models.
Yes, provider performance can vary over time due to infrastructure changes, load balancing, and updates. We continuously benchmark all providers and display historical performance trends in the "Over Time" charts.
When choosing a model on SambaNova, consider: intelligence (for quality-sensitive tasks), output speed (for throughput-intensive tasks), latency (for interactive applications requiring quick first responses), pricing (for cost-sensitive workloads), and features like context window size, JSON mode, or function calling support.