vs
Cerebras vs Dolphin
Cerebras runs open models at the fastest published speeds on wafer-scale chips. Dolphin runs one open model on idle consumer GPUs at very low prices.
By The Subconscious Team · Updated
Cerebras vs Dolphin: key differences
Cerebras lists GPT-OSS 120B near 3,000 tokens per second at $0.35 in and $0.75 out, has OpenAI as an anchor customer and trades on Nasdaq. Dolphin's API launched in October 2026 with one model, DeepSeek V4.1 Flash, at $0.01 in and $0.07 out per million, served on idle consumer GPUs. Because nodes sit across the internet, latency and uptime are less predictable, and Dolphin publishes no speed, context or SLA.
These sit at opposite ends. Cerebras is for user-facing work where generation is the wait. Dolphin is for background jobs where cost matters and a slow response does not, and for products that want uncensored models. Its wallet and Telegram signup will not suit most companies.
What Cerebras and Dolphin do
Cerebras
Cerebras builds a single chip the size of a silicon wafer, and its inference cloud is the fastest public host on the models it serves. Its developer table lists GPT-OSS 120B near 3,000 tokens per second, about six times Groq on the same weights, at $0.35 in and $0.75 out per million. The public shared catalog is thin, just GPT-OSS 120B and Qwen 3.8 27B since Gemma 4 31B moved to dedicated-only in September 2026. More model families live on dedicated endpoints and through partners like OpenRouter, Hugging Face, Vercel and AWS Marketplace.
Example models: GPT-OSS 120B, Qwen 3.8 27B
Full Cerebras profile
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.
Example models: DeepSeek V4.1 Flash, Dolphin Mistral 24B
Full Dolphin profileShould you choose Cerebras or Dolphin?
Cerebras vs Dolphin at a glance
| Attribute | Cerebras |  Dolphin |
|---|---|---|
| Model access | Open weights | Open weights, uncensored fine-tunes |
| Flagship models | GPT-OSS 120B, Qwen 3.8 27B | DeepSeek V4.1 Flash, Dolphin Mistral 24B |
| Speed | ~3,000 tok/s on GPT-OSS 120B | Unknown |
| Price | $0.35 in, $0.75 out (GPT-OSS 120B) | $0.01 in, $0.07 out (DeepSeek V4.1 Flash) |
| Customization | Unknown | Own uncensored fine-tunes |
| Deployment | Shared API, dedicated, partners | Decentralized serverless API |
| Long context | 131K on GPT-OSS 120B, 128K on Qwen 3.8 27B | Unknown |
Frequently asked questions
What is the difference between Cerebras and Dolphin?
Cerebras runs open models at the fastest published speeds on wafer-scale chips. Dolphin runs one open model on idle consumer GPUs at very low prices.
When should I choose Cerebras over Dolphin?
Fastest published output speed; Voice and live code autocomplete; Streaming UIs where generation is the wait.
When should I choose Dolphin over Cerebras?
Very cheap DeepSeek V4.1 Flash calls; Background jobs with loose latency needs; Uncensored models.
Is Cerebras or Dolphin cheaper?
Cerebras: $0.35 in, $0.75 out (GPT-OSS 120B). Dolphin: $0.01 in, $0.07 out (DeepSeek V4.1 Flash). The cheaper choice depends on the model and workload.
Related comparisons
Running long-horizon agents?
If your agents run past 200K tokens, compare both against Subconscious. Our inference stack treats a long-horizon trace as the primary workload, so speed, cost, and accuracy hold up deep into the trace.