Update README.md

README.md

@@ -33,11 +33,11 @@ For accessing Inkling via Tinker: You can get started by referring to the Tinker
 
 Inkling supports local deployment using the following open-source libraries:
 
-* SGLang ([docs](https://docs.sglang.io/cookbook/autoregressive/ThinkingMachines/Inkling), [PR](https://github.com/sgl-project/sglang/pull/31358))
+* SGLang ([recipe](https://docs.sglang.io/cookbook/autoregressive/ThinkingMachines/Inkling), [PR](https://github.com/sgl-project/sglang/pull/31358))
 * vLLM ([docs](https://recipes.vllm.ai/thinkingmachines/inkling), PR)
 * TokenSpeed ([docs](https://lightseek.org/tokenspeed/recipes/models#Inkling), [PR](https://github.com/lightseekorg/tokenspeed/pull/689))
-* Unsloth ([docs](https://unsloth.ai/docs/models/inkling), PR)
-* Huggingface ([docs](hf.co/blog/thinkingmachines-inkling), PR)
+* Unsloth ([docs](https://unsloth.ai/docs/models/inkling), [PR](https://github.com/ggml-org/llama.cpp/pull/25731))
+* Huggingface ([docs](https://hf.co/blog/thinkingmachines-inkling), PR)
 
 API access is also available through third party inference providers.
 
@@ -73,7 +73,7 @@ Training data includes a broad variety of content types, including text, images,
 
 The training data curation process includes cleaning, processing, and modifying datasets. These processing steps, which vary by data type, may include deduplication and filtering to remove junk or other low-quality data, or to advance safety or other objectives.
 
-# 5. Evaluations
+## 5. Evaluations
 
 Inkling results are reported at effort=0.99. Comparison scores are generated Jul 14, 2026. Nemotron 3 Ultra, Kimi K2.5, Kimi K2.6, GLM 5.2, and DeepSeek V4 Pro are open weights models; Gemini 3.1 Pro, Claude Fable 5, and GPT 5.6 Sol are closed weights models.
 
@@ -113,7 +113,7 @@ Inkling results are reported at effort=0.99. Comparison scores are generated Jul
 |     | StrongREJECT | 98.6%   | 98.7%            | 99.5%     | 99.8%     | 98.5%   | 98.6%           | 98.0%                 | 98.7%                | 98.5%               |
 
 
-# 6. Safety
+## 6. Safety
 
 We conducted safety evaluations ahead of release, spanning both everyday human-AI interaction and dangerous-capability testing. Because Inkling is multimodal, we paid attention to whether safety behavior held consistently across text, audio, and image inputs. We applied mitigations to reduce risks before release.
 
@@ -123,7 +123,7 @@ Across all areas, we concluded that Inkling did not present risk of material upl
 
 The residual risks identified in our evaluations — specifically, Inkling's occasional tendency to comply with role-play and indirectly framed prompts concerning harmful topics — are consistent with what you would see from any open-weight model, and are best addressed with defense-in-depth rather than relying on the model's refusals alone. Common downstream moderation tools, such as Llama Guard, are compatible with Inkling and can be layered around the model to catch jailbreak attempts, filter unsafe outputs, and enforce use-case-specific policies. We would encourage treating this kind of input/output classification as a part of your deployment stack, especially for consumer-facing or high-traffic applications where adversarial prompting is more likely.
 
-# 7. Bias, risks and limitations
+## 7. Bias, risks and limitations
 
 Inkling may exhibit general limitations common to foundation models, including hallucination (generating plausible but factually incorrect or unsupported content), occasional failures to follow instructions precisely, and degraded performance in long multi-turn conversations. As with other large-scale models trained on web-derived and synthetic data, Inkling may reflect biases present in its training data, including demographic, cultural, or linguistic biases, and may perform unevenly across languages, dialects, or subject domains that were less represented during training.