Update README.md
Browse files
README.md
CHANGED
|
@@ -45,6 +45,16 @@ Empirical results demonstrate the effectiveness of this interactive scaling. Per
|
|
| 45 |
- Handles up to 600 tool calls per task — a substantial improvement over previous open-source research agents.
|
| 46 |
- Released in 8B, 30B, and 72B parameter scales, accompanied by a comprehensive suite of tools and workflows to flexibly support diverse research settings and compute budgets.
|
| 47 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 48 |
MiroThinker v1.0 demonstrates strong general-research performance across a broad range of benchmarks, achieving 37.7%, 47.1%, 55.6%, and 81.9% on HLE-Text, BrowseComp, BrowseComp-ZH, and GAIA-Text-103, respectively. These results surpass previous open-source agents and narrow the gap with commercial counterparts such as GPT-5-high.
|
| 49 |
|
| 50 |
|
|
|
|
| 45 |
- Handles up to 600 tool calls per task — a substantial improvement over previous open-source research agents.
|
| 46 |
- Released in 8B, 30B, and 72B parameter scales, accompanied by a comprehensive suite of tools and workflows to flexibly support diverse research settings and compute budgets.
|
| 47 |
|
| 48 |
+
<div align="center">
|
| 49 |
+
|
| 50 |
+
| Model Name | Base Model | Max Length | Max Tool Calls | HF Link |
|
| 51 |
+
|:--------------------:|:---------------------------:|:----------:|:--------------:|:------------------------------------------------------------------:|
|
| 52 |
+
| MiroThinker-v1.0-8B | Qwen3-8B | 256K | 600 | [🤗 link](https://huggingface.co/miromind-ai/MiroThinker-v1.0-8B) |
|
| 53 |
+
| MiroThinker-v1.0-30B | Qwen3-30B-A3B-Thinking-2507 | 256K | 600 | [🤗 link](https://huggingface.co/miromind-ai/MiroThinker-v1.0-30B) |
|
| 54 |
+
| MiroThinker-v1.0-72B | Qwen2.5-72B-Instruct | 256K | 600 | [🤗 link](https://huggingface.co/miromind-ai/MiroThinker-v1.0-72B) |
|
| 55 |
+
|
| 56 |
+
</div>
|
| 57 |
+
|
| 58 |
MiroThinker v1.0 demonstrates strong general-research performance across a broad range of benchmarks, achieving 37.7%, 47.1%, 55.6%, and 81.9% on HLE-Text, BrowseComp, BrowseComp-ZH, and GAIA-Text-103, respectively. These results surpass previous open-source agents and narrow the gap with commercial counterparts such as GPT-5-high.
|
| 59 |
|
| 60 |
|