Update README.md
Browse files
README.md
CHANGED
|
@@ -49,6 +49,9 @@ Chat with Hermes in Nous Chat: https://chat.nousresearch.com
|
|
| 49 |
|
| 50 |
Training highlights include a newly synthesized post-training corpus emphasizing verified reasoning traces, massive improvements in math, code, STEM, logic, creativity, and format-faithful outputs, while preserving general assistant quality and broadly neutral alignment.
|
| 51 |
|
|
|
|
|
|
|
|
|
|
| 52 |
## What’s new vs Hermes 3
|
| 53 |
|
| 54 |
- **Post-training corpus**: Massively increased dataset size from 1M samples and 1.2B tokens to **~5M samples / ~60B tokens** blended across reasoning and non-reasoning data.
|
|
|
|
| 49 |
|
| 50 |
Training highlights include a newly synthesized post-training corpus emphasizing verified reasoning traces, massive improvements in math, code, STEM, logic, creativity, and format-faithful outputs, while preserving general assistant quality and broadly neutral alignment.
|
| 51 |
|
| 52 |
+
**This is the FP8 version of Hermes 4, please see the <a href="https://huggingface.co/NousResearch/Hermes-4-70B"> BF16 Model </a> if looking for that.
|
| 53 |
+
|
| 54 |
+
|
| 55 |
## What’s new vs Hermes 3
|
| 56 |
|
| 57 |
- **Post-training corpus**: Massively increased dataset size from 1M samples and 1.2B tokens to **~5M samples / ~60B tokens** blended across reasoning and non-reasoning data.
|