General-purpose coding agent · Mixture of Agents

A panel of experts beats one genius.

One prompt. Several available models think in parallel, then a team of agents verifies the work before it ships. Out comes one answer with the quality of Fable 5 and GPT-5.6— the frontier models you can't get access to right now — while staying faster and competitively priced.

A general-purpose LLM router for coding agents — across frontier and open-source models, not an OpenRouter reseller: a proprietary mixture-of-agents with its own test-verification loop. Built and battle-tested inside Speechify; now open to everyone.

Download for macOS Read documentation

Available for macOS, Linux, and Windows.

Get Waymarkcurl -fsSL https://waymark.speechify.ai/install | bash

Or read the documentation

The mechanism

The reference models are the panel.
The aggregator is the chair.

One question goes in. Several models think privately, in parallel. A sharp chair reads every answer and returns the best combined one — at the latency of the fast models, not the slow ones.

1 prompt → many models think → a team of agents verifies the work → 1 answer, better than any alone.
Opus & GPT join the panel only when the task earns the spend.

Watch it route

You can see the panel
think.

The two complaints about routed models — they're slow, and they're a black box — are exactly what Waymark fixes. Fast primaries on Cerebras and Baseten keep latency low; rare escalations only fire when a task earns them; and the routing is printed to your terminal as it happens.

Sub-5 mintasks stay sub-5 min: And finish faster than a lone Opus or Codex call when the frontier isn't needed.
Long tasksbeat the frontier: When a job is genuinely hard, the panel out-reasons any single model on the board.
Every hopvisible as it happens: No black box. Watch each model, each escalation, each handoff — live and replayable.

waymark — agent

routing live

$ waymark run "harden auth session handling"

→planharden auth session handlingGLM-5.2Baseten1.6s

→editsrc/auth/session.tsGPT-OSS-120BCerebras2.1s

→editsrc/auth/tokens.tsGPT-OSS-120BCerebras1.9s

→testpnpm test authGLM-5.2Baseten2.4s

review needs deeper reasoning — escalating GLM-5.2 → GPT-5.5 (Codex) · rare

→reviewdiff · 3 filesGPT-5.5escalation6.3s

panel agreed · 1 answer returned · you saw every hop14.3s

Why Waymark

Stop chasing the model.
Build the system around it.

A mix of today's models beats the best single model you can no longer get — with no waiting, no gated access, and none of the latency that made the first wave of orchestrators painful to use. It's not an OpenRouter-style reseller, either: Waymark is a proprietary mixture-of-agents with its own test-verification loop that checks the work before it ships back to you.

Mixture of Agents was slow. Waymark isn't.

Primaries run on Cerebras and Baseten, not on a frontier API queue. The whole panel returns before a single Opus call clears its throat.

~4smedian panel, simple task

Frontier answers, or a GLM-class bill — your pick.

Waymark Lite stays on cheap open models: ~5× cheaper than Opus, never escalates. Full Waymark adds a frontier tiebreaker when the work earns it — you choose the trade-off.

~5×cheaper than Opus — Waymark Lite

No black box. A glass one.

Every model that ran, every vote it cast, every escalation it triggered — all logged and replayable. The opacity people hate in routed models, gone.

100%of agent calls inspectable

DeepSWE Benchmark

Pass@1 on DeepSWE · higher is better · independent eval in progress

Waymark (target)FrontierOpen weights

~68

target

Waymark

MoA · GLM + GPT-OSS

target · in eval

$1.20

Claude Code

Fable 5

max · gated

—

Codex

GPT-5.5

xhigh

$9.40

Codex

GPT-5.5

medium

$4.10

Claude Code

Opus 4.8

max

$11.80

Claude Code

Opus 4.8

medium

$6.20

Opencode

Opus 4.7

medium

$3.90

Cursor CLI

GPT-5.5

medium

$3.40

Claude Code

GLM-5.2

—

$1.10

Claude Code

GLM-5.1

—

$0.90

Gemini CLI

Gemini 3.1 Pro

high

$2.80

Measured today: 95.5% on our self-verifying math litmus — the fused panel beats the strongest single model we tested by +13.6 points. The full DeepSWE run is in independent evaluation now; the bar above is our target, and Fable itself is gated to 100 partners.

Run it yourself

Math-litmus result: 22 self-verifying integer-answer problems, gold checked by brute force (fusion 95.5% vs strongest single model 81.8%; small, directional sample). DeepSWE competitor bars are illustrative and the Waymark bar is a target — independent DeepSWE eval in progress.

Pricing

Two ways to run Waymark.

Pick the trade-off you actually want. Lite stays on cheap open-weights and never touches a frontier model — the lowest, most predictable bill. Waymark escalates to a frontier model when a turn is hard, so you get frontier-grade quality from a mixture that beats any single model.

Waymark Lite

Cheapest

Open-weights only. Never calls a frontier model.

$0.40/ M in · $1.44 / M out

Flat open-weights rate — no frontier surcharge, ever.

GLM-5.2 · GPT-OSS-120B

—~5× cheaper than an Opus-max run
—Never escalates — predictable, flat pricing
—A mixture of open models still beats any single open model

Waymark

Best quality

Frontier when it counts — a mixture that beats any single model.

frontier/ M in · rates / M out

Open-weights on easy turns; frontier rates on the hard ones.

GLM-5.2 · GPT-OSS-120B → Opus 4.8 / GPT-5.5

—Escalates to a frontier model on hard or ambiguous turns
—The panel + a frontier tiebreaker beats a lone frontier model
—You're buying quality, not a discount — hard work bills near frontier

On the roadmap: per-model spend & usage caps — set a ceiling on Opus, GPT, or any model, per task or per workspace.

“~5× cheaper” = measured median of Opus-max ÷ Waymark Lite cost across a coding-query sample (range 1.7–7.2×). Waymark's blended cost depends on how often a workload escalates; frontier-heavy work (e.g. complex code) trends toward frontier rates — that's the quality you're paying for, not a billing surprise.

One command

The model is the part you swap.
The system is the thing you own.

Flip the whole panel on with one command. Pick a preset, or pin your own reference models and aggregator. Provider-agnostic — plug in Baseten, Cerebras, Anthropic, OpenAI, or your own endpoint.

Drop-in: keeps your existing agent loops and tools
Per-surface routing: default, review, plan
Swap a reference model without touching your workflow

waymark — zsh

$ waymark moa

# turn the panel on

$ waymark model default --provider moa

# route normal turns through the panel

$ waymark model review --provider moa

# use the panel only for reviews

$ waymark presets

# fast-pair · opus-aggregator · budget

panel: glm-5.2 + gpt-oss-120b → aggregator

escalation: opus-4.8 (armed, idle)

✓ mixture of agents active

No harness required

Already have an agent?
Use the panel as a model.

Don't want to use our harness? Point any OpenAI-compatible client at Waymark and the whole mixture of agents answers behind a single model id. Swap your base_url and model — nothing else changes.

waymark-moabalanced panel · auto-escalates
waymark-fastprimaries only · never escalates
waymark-maxaggressive escalation for hard tasks

Routing decisions come back in x-waymark-route response headers, so the panel stays transparent even over the API.

api.speechify.ai — chat/completions

curl --location 'https://api.speechify.ai/v1/chat/completions' \
  --header 'Content-Type: application/json' \
  --header "Authorization: Bearer ${WAYMARK_API_KEY}" \
  --data '{
    "model": "waymark-moa",
    "max_completion_tokens": 1024,
    "temperature": 0.2,
    "top_p": 1,
    "stream": false,
    "reasoning_effort": "medium",
    "messages": [
      { "role": "user", "content": "Why is fast inference important?" }
    ]
  }'

Waymark is fast because of the silicon underneath it.

Served on the fastest inference in AI — on purpose.

Cerebras

wafer-scale inference

Primary · always on

Baseten

dedicated GLM endpoints

Primary · always on

Anthropic

Opus escalation

On escalation

OpenAI

GPT escalation

On escalation

The winning move isn't waiting.
It's the panel.

Install Waymark and point your agent at the panel. Frontier-class output, fast, today — no gated access required.

Download for macOS Read documentation

Available for macOS, Linux, and Windows.

A panel of experts beats one genius.

The reference models are the panel. The aggregator is the chair.

You can see the panelthink.

Stop chasing the model.Build the system around it.

Mixture of Agents was slow. Waymark isn't.

Frontier answers, or a GLM-class bill — your pick.

No black box. A glass one.

DeepSWE Benchmark

Two ways to run Waymark.

Waymark Lite

Waymark

The model is the part you swap.The system is the thing you own.

Already have an agent?Use the panel as a model.

Served on the fastest inference in AI — on purpose.

The winning move isn't waiting.It's the panel.

The reference models are the panel.
The aggregator is the chair.

You can see the panel
think.

Stop chasing the model.
Build the system around it.

The model is the part you swap.
The system is the thing you own.

Already have an agent?
Use the panel as a model.

The winning move isn't waiting.
It's the panel.