70.55B Llama scale
Captured config identifies LlamaForCausalLM and Safetensors metadata reports 70553804800 parameters.
Open Source Model Profile · mattshumer
Reflection-Llama-3.1-70B
Reflection-Llama-3.1-70B is a 70.55B-parameter Llama-family text-generation fine-tune from mattshumer. Its model card documents Reflection-Tuning, thinking and output tags, and a dedicated reflection system prompt.
- Publisher
- mattshumer
- Task
- text-generation
- Model type
- llama
- License
- llama3.1
- Library
- transformers
- Publication status
- Accepted · not indexed
Model overview
Reflection-Llama-3.1-70B is published by mattshumer as a Llama-based text-generation model. The captured configuration identifies LlamaForCausalLM and Safetensors metadata reports 70553804800 parameters. Hub tags associate it with meta-llama/Llama-3.1-70B-Instruct as base, while the model card describes Reflection-Tuning on synthetic Glaive data.
Recorded capabilities
Reflection-Tuning description
The model card describes Reflection-Tuning as teaching the model to detect mistakes in its reasoning and correct course.
Thinking and output tags
According to the model card, reasoning appears inside trained <thinking> tags and final answers inside trained <output> tags, with <reflection> tags for correction.
Documented system prompt
The publisher recommends an exact system prompt for reasoning, output formatting, and in-reasoning self-correction.
Llama 3.1 Instruct lineage tags
Hub tags list meta-llama/Llama-3.1-70B-Instruct as base model and fine-tune, and the card says it samples like any other Llama model.
Use cases in the source record
- Reasoning-style text generation that uses the documented thinking, output, and reflection tags.
- Llama 3.1 chat-template experiments that apply the publisher's recommended reflection system prompt.
Limitations and unknowns
- No evaluation results were extracted from this record.
- No context-window value was extracted from this record.
- Provider state is historical snapshot data, not independently refreshed current availability.
- The publisher reports an initial upload issue it believes was fixed, and describes a future dataset, report, and 405B model as plans; none of this has been independently verified by Ethen.
Source and provenance
Source: mattshumer/Reflection-Llama-3.1-70B
Captured: Unknown. Processed: 2026-09-07T19:34:51.395207+00:00.
Reflection Llama-3.1 70B | IMPORTANT UPDATE – There was an issue with the model when we first uploaded it. If you tried it and didn't have good results, please, try again, we think we've fixed the issue. Reflection Llama-3.1 70B is an open-source LLM, trained with a new technique called Reflection-Tuning that teaches a LLM to detect mistakes in its reasoning and correct course. The model was trained on synthetic data generated by Glaive . If you're training a model, Glaive is incredible — use them. You can try the model here . Benchmarks Trained from Llama 3.1 70B Instruct, you can sample from Reflection Llama-3.1 70B using the same…
F001F002F003F004F005F006F007F009F010F011F012F015F016F017F018F019