SSyncDev

SenseNova 6.8 Code2Video

SyncDev model directory · Profile updated Nov 18, 2026

Compare with another model

SenseNova 6.8 Code2Video is SenseTime's open 27B model that makes short motion-design videos by writing and rendering executable code.

Key specifications

Specification score
58.4
Context window
524.29K
License
Apache-2.0

The specification score summarizes recorded features and limits. It is not a benchmark result.

Pricing conditions: No API pricing stated; open weights for self-hosting

Inputs and outputs

Recorded modality support shows what this model can accept and produce. An undocumented modality is shown as unknown rather than assumed to be unsupported.

Accepts

  • Text input
  • Image input

Produces

  • Text output

Capabilities

ReasoningFunction callingVisionOpen weights

What is SenseNova 6.8 Code2Video?

SenseNova 6.8 Code2Video does not paint pixels. It writes programs, and the programs render the video. That sounds like a detour until you need to change one word in a title card, retime a transition or keep a brand colour exact, all things that are trivial in code and painful in a diffusion model.

SenseTime released it on 24 October 2026 as a lightweight open model fine-tuned from Qwen3.8-27B. The current focus is narrow on purpose: short videos implemented entirely in code, where composition and motion choreography matter more than photorealism.

Why code as the medium

Code is editable, reviewable and deterministic. It can also call other tools. SenseTime frames the longer-term goal as a model that coordinates image, video, 3D and editing tools through code, chaining generation and compositing into one workflow. This release is the first step, not the finished system.

What ships

Resource Purpose
Model weights sensenova/SenseNova-6.8-Code2Video, about 27.4B parameters
trajectories/ Message and tool logs, with figures, from real runs
Production skill Workflow, craft guides and rendering tools
Licence Apache-2.0

Serving it

Both reference recipes assume four GPUs with tensor parallelism and a very long request budget of 524,288 tokens, which reflects how long these agentic, tool-calling sessions get. The vLLM recipe enables the Qwen3 reasoning parser, so the model thinks before it writes code. LightLLM is the other documented path.

Who should try it

  • Teams producing explainer clips, product animations or data stories where every frame must be revisable.
  • Developers building agent workflows that already render with code and want a model tuned for that loop.
  • Anyone comparing code-driven creation with prompt-to-video systems in the text-to-video directory.

A fair evaluation

Give it ten storyboards with exact text, timing and colour requirements. Count how many render without a manual fix, then make one change request per video and check whether the edit stays local. That second number tells you more than the first. Compare it with a general coding model from the code generation directory given the same skill and tools, to see how much the fine-tune actually buys you.

How to evaluate SenseNova 6.8 Code2Video

Compare the model on the work you need to complete.

Tasks and inputs

SenseNova 6.8 Code2Video is listed for Text-to-video, Coding & Code Generation. It accepts text, image and produces text. Use the publisher’s input format and preprocessing for your chosen deployment.

Capacity and cost

Documented limits: 524.29K context. Measure a completed task using the prompts, documents and tools your application will use. Include reasoning and retries in the result.

Choosing a deployment

The weights are downloadable under Apache-2.0. Check runtime support and size the complete model and working context for your expected concurrency.

Frequently asked questions

It is fine-tuned from Qwen3.8-27B and has roughly 27.4B parameters.

These records share a directory category. Check the source card, output type and deployment requirements before treating a model as a replacement.

Thinkingmachines

Inkling 2 Mini

Inkling 2 Mini is Thinking Machines' Apache 2.0 looped MoE model (26.4B total, 3.4B active) for agentic coding that runs on a 32 GB laptop.

1M context

radixark

DaisyRose-27B

DaisyRose-27B is RadixArk's Apache 2.0, RL-only post-train of Qwen3.8-27B for agentic coding, with a 256K context and image input.

262.14K context

Microsoft

MAI Code 2

MAI Code 2 is Microsoft AI's coding model for multi-file work, debugging and frontend builds, live in GitHub Copilot at $0.50/$2.50 per 1M (promo).

— context

Microsoft

MAI Code 2 Flash

MAI Code 2 Flash is Microsoft AI's low-cost coding model for edits, fixes and explanations, on every GitHub Copilot plan including Free and Student.

— context

Microsoft

MAI Cyber 2 Flash

MAI Cyber 2 Flash is Microsoft AI's security model for vulnerability discovery and threat investigation, with a 1M-token context, shipping in MDASH.

1M context

Google

Gemini 4 Hydrogen

Gemini 4 Hydrogen is Google's fast, lower-cost Gemini 4 model for everyday coding and agentic execution, with a 1M context, xhigh thinking and a free tier.

1.05M context