# SambaNova vs Dolphin

> SambaNova sells premium speed on custom chips. Dolphin sells very low prices on idle consumer GPUs.

Canonical: https://www.subconscious.dev/compare/sambanova-vs-dolphin · By The Subconscious Team · Updated October 11, 2026

## How they compare

SambaNova serves MiniMax M3, MiniMax M2.7, DeepSeek, Gemma 4 31B and GPT-OSS 120B on its own RDU chips, with GPT-OSS 120B at $0.22 in and $0.59 out and fast decode as the selling point. Dolphin launched its API in October 2026 with one model, DeepSeek V4.1 Flash, at $0.01 in and $0.07 out, run across consumer GPUs on the internet with no published speed, context or SLA.

SambaNova fits interactive coding agents and copilots that need fast tokens on big open models. Dolphin fits background calls where latency matters little, and products that need uncensored models. Its wallet-and-Telegram signup will not clear most procurement teams.

## What each one does

### SambaNova

SambaNova designs its own inference chip, the Reconfigurable Dataflow Unit, and sells fast tokens on large open models through SambaCloud. The RDU maps the model graph onto the chip to cut trips to off-chip memory. A three-tier memory design of SRAM, HBM and bulk DRAM lets one system host very large models and hot swap between several of them in milliseconds. SambaCloud serves models like MiniMax M3 (in preview with a 1M context), MiniMax M2.7, DeepSeek, Gemma 4 31B and GPT-OSS 120B, with speeds reported by Artificial Analysis.

### Dolphin

Dolphin began in 2023 as Eric Hartford's series of open-source, uncensored fine-tunes, trained to drop the refusals and baked-in guidelines of their base models so that the developer sets the rules through the system prompt. Current releases include Dolphin Llama 3.1 8B and Dolphin Mistral 24B Venice Edition, built with Venice, and the project publishes nearly 100 models on Hugging Face. It also runs a chat app, a Telegram bot and a synthetic data service that generates answers to commonly refused prompts.

## Which is best, and when

### Choose SambaNova for

- Fast decode on large open models
- Interactive coding agents and copilots
- Millisecond model hot swapping

### Choose Dolphin for

- Very cheap DeepSeek V4.1 Flash calls
- Models without vendor-imposed refusals
- Latency-tolerant background jobs

## At a glance

| Attribute | SambaNova | Dolphin |
|---|---|---|
| Model access | Open weights | Open weights, uncensored fine-tunes |
| Flagship models | MiniMax M3, MiniMax M2.7, GPT-OSS 120B | DeepSeek V4.1 Flash, Dolphin Mistral 24B |
| Speed | ~800 tok/s on MiniMax M2.7 (SN50, SemiAnalysis) | - |
| Price | $0.22 in, $0.59 out (GPT-OSS 120B) | $0.01 in, $0.07 out (DeepSeek V4.1 Flash) |
| Customization | - | Own uncensored fine-tunes |
| Deployment | SambaCloud, racks for neoclouds | Decentralized serverless API |
| Long context | Up to 1M (MiniMax M3, preview) | - |

## FAQ

### What is the difference between SambaNova and Dolphin?

SambaNova sells premium speed on custom chips. Dolphin sells very low prices on idle consumer GPUs.

### When should I choose SambaNova over Dolphin?

Fast decode on large open models; Interactive coding agents and copilots; Millisecond model hot swapping.

### When should I choose Dolphin over SambaNova?

Very cheap DeepSeek V4.1 Flash calls; Models without vendor-imposed refusals; Latency-tolerant background jobs.

### Is SambaNova or Dolphin cheaper?

SambaNova: $0.22 in, $0.59 out (GPT-OSS 120B). Dolphin: $0.01 in, $0.07 out (DeepSeek V4.1 Flash). The cheaper choice depends on the model and workload.

## Running long-horizon agents?

If your agents run past 200K tokens, compare both against Subconscious: [Subconscious vs SambaNova](https://www.subconscious.dev/compare/subconscious-vs-sambanova.md), [Subconscious vs Dolphin](https://www.subconscious.dev/compare/subconscious-vs-dolphin.md).

Full profiles: [SambaNova](https://www.subconscious.dev/providers/sambanova.md), [Dolphin](https://www.subconscious.dev/providers/dolphin.md).