Cognitive Computations
12 models across 1 family · Latest: Dolphin 2.9.2 Qwen2-72B (2024-05)
Uncensored AI models for open access
Cognitive Computations's portfolio covers 12 active models across 1 current family, spanning coding, rag, and agents. Open a model detail page to compare provider routes and sourced benchmarks.
Covers 6 workload areas across 12 active tracked models; last verified 2026-05-19.
Use it for
- Teams evaluating coding, rag, and agents across this lab's releases
- Comparing model families before committing to a flagship
- Migration and pricing follow-ups across 12 tracked models
Do not use it for
- Choosing a hosting provider without opening a model page for price ladders
Active models
12
Current models from this lab, excluding deprecated ones
Active families
1
Current model families from this lab
Open catalog
12 open
12 open source / 0 open weights
Lowest output price
$0.300 /1M
Cheapest tracked output across active models, per 1M tokens
Latest dated release
2024-05-27
Dolphin 2.9.2 Qwen2-72B
Freshness
2026-05-19
Researched 106d ago
Information
Release cadence
- Dated releases
- 5
- Latest model
- Dolphin 2.9.2 Qwen2-72B
- Latest date
- 2024-05-27
Where this lab wins
- Coding: 2 tracked models with SWE-bench / HumanEval-style scores.
- RAG: 1 tracked model with ruler / needle retrieval benchmarks.
- Agentic: 1 tracked model with BFCL, tau-bench, and SWE-bench tool-use coverage.
- Long-context: 3 tracked models with context-token or InfiniteBench-class signal.
Flagship quality / price signal
- Flagship
- Dolphin 2.6 Mixtral 8x7B
- Selection source
- Best sourced coding Q/$
- Coding grade
- unknown
- Benchmark / output price
- Not enough sourced benchmark and price coverage
About
Cognitive Computations is the research group behind the Dolphin model lineage, with Eric Hartford and collaborators publishing open-weight instruction fine-tunes for chat, coding, function calling, and agent-style workflows. Its Dolphin releases are best known for deliberately reducing alignment filtering in the training data, which makes the models highly steerable but shifts moderation, policy controls, and deployment safeguards to the application owner. The group publishes model cards, datasets, and checkpoints through the Dolphin Hugging Face organization and related open-source channels.
Featured models
| Model | Released | Context | Input price ($/1M) | Output price ($/1M) | License | Openness |
|---|---|---|---|---|---|---|
| Dolphin 2.9.2 Qwen2-72B | 2024-05-27 | 128k | $0.90 | $0.90 | Apache 2.0 | Open source |
| Dolphin 2.6 Mixtral 8x7B | 2023-12-18 | 32k | $0.15 | $0.45 | Apache 2.0 | Open source |
| Dolphin 2.5 Mixtral 8x7B | 2023-12-18 | 32k | $0.6 | $0.6 | Apache 2.0 | Open source |
Model families
Recent releases
- Dolphin 2.9.2 Qwen2-72B- 2024-05-27
- Dolphin 2.6 Mixtral 8x7B- 2023-12-18
- Dolphin 2.5 Mixtral 8x7B- 2023-12-18
- Dolphin 2.9 Llama 3 8B- 2023-12-18
- Dolphin 2.7 Mixtral 8x7B- 2023-12-18
Explore related pages
Last reviewed: 2026-05-19. Data sourced from public lab announcements and provider documentation.
