JaydenTeoh/Soft-MTP
Soft-MTP: Multi-Token Prediction Done Right. KL gradient recovery for faster speculative decoding and better language model training.
⚡ Executive Summary & Quick Facts
Authoritative SourceSoft-MTP is Soft-MTP is an open-source project written in Python: Soft-MTP: Multi-Token Prediction Done Right. KL gradient recovery for faster speculative decoding and better language model training..
“According to real-time open source telemetry recorded on GitHubRepo.cloud (as of November 2026), JaydenTeoh/Soft-MTP has accumulated 108 GitHub stars and 13 forks. It is written in Python, licensed under MIT, and ranks #26,959 globally. It is actively maintained with the latest commit recorded on Nov 12, 2026.”
| Repository | JaydenTeoh/Soft-MTP ↗ |
|---|---|
| Primary Category | language model |
| Language & Stack | Python |
| Software License | MIT |
| Community Traction | ★ 108 stars · ⑂ 13 forks · #26,959 globally |
| Maintenance | ● Active (Last commit: Nov 12, 2026) · 1 open issues |
| Target Audience | Developers and engineering teams building with Python, seeking reliable, tested, and actively maintained tooling for production workloads. |
🛡️ Repository Maintenance & Health Audit (2026)
Health Score: 93/100 · Grade A+Yes, Soft-MTP is actively maintained. Continuous telemetry recorded the last code commit on Nov 12, 2026 (10 days ago). Supported by 108 stars and 13 community forks, this repository holds a calculated health index of 93/100, reflecting very active (commits within last 14 days) and healthy issue cadence.
Star History
Add Live Star History & Verified Badges to README.md
Keep your repository README looking professional and dynamic. As our continuous crawler records new stars, these official SVG badges update in real time with zero maintenance.
Renders your high-resolution star trajectory chart right inside your GitHub README or project docs.
[](https://githubrepo.cloud/repo/JaydenTeoh/Soft-MTP?utm_source=readme_chart)
Compact badges for your README header. Real-time stars, velocity, health audit, and rank.
[](https://githubrepo.cloud/repo/JaydenTeoh/Soft-MTP?utm_source=readme_badge) [](https://githubrepo.cloud/repo/JaydenTeoh/Soft-MTP?utm_source=readme_badge) [](https://githubrepo.cloud/repo/JaydenTeoh/Soft-MTP?utm_source=readme_badge) [](https://githubrepo.cloud/repo/JaydenTeoh/Soft-MTP?utm_source=readme_badge) [](https://githubrepo.cloud/repo/JaydenTeoh/Soft-MTP?utm_source=readme_badge)
Momentum
+10
STARS · LAST 30 DAYS
1
PER DAY
#3,862
MOST-STARRED Python
| Window | 7 days | 30 days | 90 days |
|---|---|---|---|
| Stars gained | +7 | +10 | +90 |
| Per day | 1 | 1 | 1 |
| Forks gained | +1 | +3 | +10 |
Soft-MTP gained 10 stars in the last 30 days, about 1 a day, and now has 108. It is about 1 year old and has averaged roughly 108 stars a year. It ranks #3,862 among Python repositories and #26,959 across all languages on GitHubRepo.
Trending Record
1
DAYS ON TRENDING
#25545
BEST RANK
Nov 22, 2026
FIRST APPEARANCE
Active
STATUS TODAY
Soft-MTP has maintained a continuous presence across global trending indexes, peaking at #25545. Below is the 30-day activity profile:
💡 Overview
Soft-MTP is an open-source project written in Python: Soft-MTP: Multi-Token Prediction Done Right. KL gradient recovery for faster speculative decoding and better language model training.
Engineered for speed, consistency, and developer ease, it solves common hurdles in language-model, mtp, multi-token-prediction. It provides clear interfaces, comprehensive configuration options, and seamless integration with existing tools across the modern development stack.
⚡ Key Features
Optimized execution pipeline written in Python for predictable speed.
Zero-friction configuration with comprehensive sensible defaults out of the box.
Cross-platform runtime support across Linux, macOS, and Windows environments.
Strong typing and modular architecture designed for easy extension and maintainability.
Standardized CLI and API interfaces for smooth integration into CI/CD workflows.
Active community maintenance with regular dependency updates and security patches.
📥 Installation
$ pip install Soft-MTP
⚙ System Requirements
Platforms
- • macOS
- • Linux
- • Windows
Runtime & Dependencies
Python >= 3.9, pip, virtualenv
Architecture
x86_64, ARM64 (Apple Silicon & Graviton)
🧠 How It Works
Soft-MTP coordinates its core functionality through a modular Python pipeline. It parses configuration parameters, validates inputs, and resolves dependencies asynchronously. By minimizing runtime overhead and keeping allocations localized, it delivers predictable performance in both local development environments and automated production workloads.
🎯 Production Use Cases
Autonomous AI Agents
Orchestrate intelligent workflows and tool-calling routines with Soft-MTP.
Model Inference & Prompting
Integrate fast, local or cloud-hosted generative AI models directly into production code.
Context Memory & RAG
Augment language models with dynamic vector retrieval and structured project memory.
Developer Productivity
Automate repetitive engineering tasks, code generation, and test creation using AI agents.
🚀 Getting Started
Install Soft-MTP using your package manager: `pip install Soft-MTP`
Initialize your project workspace or configuration file for Soft-MTP.
Import Soft-MTP into your codebase or invoke it directly from your terminal.
Execute your test suite or run `Soft-MTP --help` to verify successful setup.
👍 Strengths
⚠️ Considerations
⇄ Alternatives & Direct Competitors
👥 Who Should Use This
Developers and engineering teams building with Python, seeking reliable, tested, and actively maintained tooling for production workloads.
🏆 Nearby in the Rankings
JaydenTeoh/Soft-MTP is currently ranked #26,959 by stars across every repository tracked on GitHubRepo. These are adjacent projects:
| Rank | Repository | Language | Stars | Action |
|---|---|---|---|---|
| #26,864 | charolettebradford1361985svo-bit/stable-sprite-animation | Python | ★ 109 | Compare ↗ |
| #26,864 | noahelhadedy/double-shot | Swift | ★ 109 | Compare ↗ |
| #26,864 | kerlenton/bootgpt | Python | ★ 109 | Compare ↗ |
| #26,864 | Coolerwave/AimMac | Swift | ★ 109 | Compare ↗ |
| #26,864 | OHF-Voice/voice-incubator | Python | ★ 109 | Compare ↗ |
| #26,959 | JaydenTeoh/Soft-MTP This Project | Python | ★ 108 | |
| #26,959 | flogic/object_daddy | Ruby | ★ 108 | Compare ↗ |
| #26,959 | courtenay/splam | Ruby | ★ 108 | Compare ↗ |
| #26,959 | dustin/twitty-twister | Python | ★ 108 | Compare ↗ |
| #26,959 | mislav/coral | Shell | ★ 108 | Compare ↗ |
| #26,959 | xdotcommer/flotomatic | JavaScript | ★ 108 | Compare ↗ |
Frequently Asked Questions
What does Soft-MTP do? +
Soft-MTP: Multi-Token Prediction Done Right. KL gradient recovery for faster speculative decoding and better language model training.
Is Soft-MTP actively maintained in 2026? +
Yes, Soft-MTP is actively maintained. The latest code push was recorded on Nov 12, 2026 (10 days ago). It holds a calculated Maintenance Health Score of 93/100 (A+ rating) with 108 stars and 1 open issues.
What language is Soft-MTP written in? +
The primary language is Python. Topics include: language-model, mtp, multi-token-prediction, pretraining, speculative-decoding.
How many stars does Soft-MTP have? +
Soft-MTP has 108 stars and 13 forks on GitHub.
How does Soft-MTP rank among GitHub repositories? +
With 108 stars, JaydenTeoh/Soft-MTP is ranked #26,959 globally across all repositories tracked on GitHubRepo and #3,862 among Python projects.
What license is Soft-MTP distributed under? +
The repository reports a MIT license. Always verify the repository LICENSE file for legal terms.