---
title: "Speculative Decoding - Air Market"
description: "Your model works but each reply takes too long, and you cannot throw more hardware at it. This covers ways of generating tokens in parallel instead of strictly one at a time, including draft models a…"
canonical: "https://airmarket.ai/app/speculative-decoding"
---

[Air Market](/) / [Data & Analytics](/category/data) / Speculative Decoding

Skill by [Orchestra Research](/@orchestra-research)

# Speculative Decoding

Get answers out of a model sooner without buying more compute

## What this does

Your model works but each reply takes too long, and you cannot throw more hardware at it. This covers ways of generating tokens in parallel instead of strictly one at a time, including draft models and lookahead decoding. Reach for it when latency is the complaint, or when the compute you have is fixed.

### Top 3 things it does for you

- Cut the wait on a real-time feature
- Run a model on the hardware you already have
- Try draft models and multiple-head decoding

## Information

**Speculative Decoding at a glance**

|  |  |
| --- | --- |
| Kind | Skill |
| Publisher | [Orchestra Research](/@orchestra-research) |
| Category | [Data & Analytics](/category/data) |
| Works with | Agent Skills - an open standard supported by 40+ agents |
| Version | 1.0.0 |
| Size | 32 KB |
| Licence | MIT |
| Repository | [orchestra-research/ai-research-skills](https://github.com/orchestra-research/ai-research-skills/tree/28f2d29236f2bade2eb92cadb2585189589a5828/19-emerging-techniques/speculative-decoding) |
| GitHub stars | 11,987 (as of Nov 2026) |

## Security report

**Verified.** AIR scanned this exact version of Speculative Decoding and flagged nothing. AIR's verdict: **safe**.

**What Speculative Decoding's check was run against**

|  |  |
| --- | --- |
| Commit | 28f2d29236f2bade2eb92cadb2585189589a5828 |
| Content hash | f860bab557ee28091a8742173dd34ea86e0d73416d56173cc3f530694c5c7f14 |
| Files read | [Browse the files at this exact commit](https://github.com/orchestra-research/ai-research-skills/blob/28f2d29236f2bade2eb92cadb2585189589a5828/19-emerging-techniques/speculative-decoding/SKILL.md) |

A scan is named by the exact content it read, not by a version number. A different version is a different artifact, and nothing here applies to a version AIR has not scanned.

[Read the full security report for Speculative Decoding](/security/speculative-decoding)

## About Speculative Decoding

Generating one token at a time is often what makes a model feel slow. This skill covers the family of techniques that break that pattern: speculative decoding with a draft model, Medusa's multiple heads, and lookahead decoding, along with the reported speedup range for each situation.

It goes into the pieces those methods rest on - draft models, tree-based attention, Jacobi iteration, parallel token generation - and into what changes when you move from an experiment to a deployment.

Use it when you are tuning inference speed, when a real-time application needs lower latency, or when the compute available to you is capped and the model still has to run.

## What's inside

- SKILL.md - instructions
- references/lookahead.md - reference
- references/medusa.md - reference

## Questions

### Is Speculative Decoding safe to install?

AIR scanned this exact version of Speculative Decoding and flagged nothing. AIR's verdict: safe. The verdict is named by the content that was read, at commit 28f2d29236f2, so it cannot drift onto a version nobody scanned. The full report is at https://airmarket.ai/security/speculative-decoding.

### How do I install Speculative Decoding?

Open https://airmarket.ai/app/speculative-decoding on Air Market and use the install action: Air Market hands Claude a prompt naming this exact skill, and Claude sets it up. Source, pinned to the exact commit this listing describes: https://github.com/orchestra-research/ai-research-skills/tree/28f2d29236f2bade2eb92cadb2585189589a5828/19-emerging-techniques/speculative-decoding. Install from that link, not the branch.

## Related on Air Market

- [More Data & Analytics listings](/category/data)
- [Everything by Orchestra Research](/@orchestra-research)

---

Canonical URL: <https://airmarket.ai/app/speculative-decoding>