docs: fix reasoning_content field name and add Video to the modality row

[Mr-Neutr0n]

Two inconsistencies inside the README. Both are checkable against the document itself — no external knowledge needed.

1. The usage example reads the wrong field

Section 6 states the model "will return reasoning_content", and the example message uses reasoning_content as its key. The print statement then reads .reasoning:

print(f"response: {response.choices[0].message.reasoning}")

Counting occurrences in that section: reasoning_content appears at lines 617, 619 and 632; .reasoning appears once, at line 649. Copy-pasting the snippet prints None instead of the preserved thinking it exists to demonstrate — which is the entire point of the example.

2. The Model Summary table omits video

The table lists:

Modality	Text, Image

But the Key Features immediately above say Kimi K3 "understands text, images, and video within the same model", and the evaluation table reports:

Video-MME (w. sub)	90.0

So the spec table and the benchmark table give a reader contradictory answers about whether video is supported. Given the model is benchmarked on video and the feature list claims it, the table row looks like the thing that's out of date.

---

Diff is 2 lines. Happy to split into two PRs if you'd prefer them reviewed separately, and equally happy to close this if the modality row is deliberate and video is served by something other than the base weights.

Thanks for shipping the weights openly.