---
type: "post"
id: "openrouter-raises-listed-token-prices-for-meta-muse-ace131"
slug: "openrouter-raises-listed-token-prices-for-meta-muse-ace131"
title: "OpenRouter raises listed token prices for Meta Muse Glimmer 30B"
description: "OpenRouter’s model page records a higher listed rate at Together for Meta Muse Glimmer 30B: $0.35 per million input tokens and $1.50 per million output tokens, versus the earlier $0.30/$1.20 headline rates."
retrieval_nugget: "OpenRouter’s model page records a higher listed rate at Together for Meta Muse Glimmer 30B: $0.35 per million input tokens and $1.50 per million output tokens, versus the earlier $0.30/$1.20 headline rates. Teams using the model through OpenRouter should distinguish the displayed rate from their routed provider mix and effective bill."
published_at: "2026-11-17"
updated_at: "2026-11-17"
record_date: "2026-11-17"
date_kind: "published_at"
topics: ["pricing","inference"]
entities: ["OpenRouter","Meta","Muse Glimmer 30B","Together","ModelRun","DeepInfra"]
related_patterns: []
novelty_label: "new_fact"
what_changed: "OpenRouter lists Meta’s Muse Glimmer 30B at $0.30 per million input tokens and $1.20 per million output tokens."
prior_refs: []
source_url: "https://openrouter.ai/meta/muse-glimmer-30b"
source_urls: ["https://openrouter.ai/meta/muse-glimmer-30b"]
schema_version: "newruntime-agent-readable-v0.2"
stable_id: "post:openrouter-raises-listed-token-prices-for-meta-muse-ace131"
status: "published"
visuals: []
editor: {"role":"editor","name":"Andrey Reshetnikov","profile":"https://newruntime.com/owner-profile.md","profile_json":"https://newruntime.com/owner-profile.json","same_as":["https://www.linkedin.com/in/reshetnikov1/","https://t.me/lapetuse"]}
routes: {"html":"https://newruntime.com/posts/openrouter-raises-listed-token-prices-for-meta-muse-ace131/","markdown":"https://newruntime.com/posts/openrouter-raises-listed-token-prices-for-meta-muse-ace131.md","json":"https://newruntime.com/posts/openrouter-raises-listed-token-prices-for-meta-muse-ace131.json"}
markdown_url: "https://newruntime.com/posts/openrouter-raises-listed-token-prices-for-meta-muse-ace131.md"
---

# OpenRouter raises listed token prices for Meta Muse Glimmer 30B

## Retrieval answer

OpenRouter’s model page records a higher listed rate at Together for Meta Muse Glimmer 30B: $0.35 per million input tokens and $1.50 per million output tokens, versus the earlier $0.30/$1.20 headline rates. Teams using the model through OpenRouter should distinguish the displayed rate from their routed provider mix and effective bill.

## Summary
OpenRouter’s model page records a higher listed rate at Together for Meta Muse Glimmer 30B: $0.35 per million input tokens and $1.50 per million output tokens, versus the earlier $0.30/$1.20 headline rates. Teams using the model through OpenRouter should distinguish the displayed rate from their routed provider mix and effective bill.

## What happened
OpenRouter’s listing describes Muse Glimmer 30B as a dense, open-weight multimodal model from Meta Superintelligence Labs. It says the 30-billion-parameter model was distilled from Muse Spark and optimized for autonomous agents on consumer hardware. The page positions it for long-horizon agentic and coding work, including multi-step reasoning, tool use, failure recovery, image understanding, and multilingual operation across more than 100 languages.

The listing gives Muse Glimmer 30B a 131,072-token context window. It accepts text and image inputs and returns text. OpenRouter says the model supports function calling through `tools` and `tool_choice`, plus structured output through a JSON schema in `response_format`. Developers can call it with the model identifier `meta/muse-glimmer-30b` through OpenRouter’s OpenAI-compatible API, including streaming and non-streaming requests. The page dates the model’s release to August 9, 2026.

The relevant change is in provider-level listed pricing. Together now shows $0.35 per million input tokens and $1.50 per million output tokens. OpenRouter’s price history identifies the prior headline rates as $0.30 per million input tokens and $1.20 per million output tokens. ModelRun, by Modular, and DeepInfra currently show the lower $0.30/$1.20 input/output rates, while all three providers show $0.04 per million cache-read tokens. This is a change in an OpenRouter provider listing; it does not establish that Meta changed a price outside that route.

OpenRouter says the same model can be served by different companies and routed according to Balanced, Nitro, or Floor modes: price and speed balanced, fastest, or cheapest. It also says users can pin or exclude providers and that its routing can recover from an upstream error by sending eligible traffic to another healthy provider. The page’s weighted average is separate from posted provider pricing: it displays $0.0662 per million input tokens and $1.351 per million output tokens. OpenRouter says caching and discounts mean customers often pay less than listed prices. Its page also displays provider latency, throughput, cache-hit, tool-call-error, structured-output-error, uptime, and availability measurements, making this a routing decision as well as a price observation.

## What the archive adds
New Runtime’s earlier coverage on October 20, 2026 described Muse Glimmer 30B as Meta’s open 30-billion-parameter model for autonomous work on consumer hardware. The present event is a new pricing fact on that known model, rather than a model launch: Together’s listed OpenRouter rate rose while ModelRun and DeepInfra continue to display lower rates.

## Trend signal
This is a concrete token-cost change for a model promoted for agent work, but it does not yet demonstrate a move to economics measured by verified completed tasks. OpenRouter provides token rates, cache statistics, provider shares, and operational measurements; it does not connect them to task success or a completed-work cost. This does not yet fit a pattern New Runtime tracks.

## Why it matters
Small unit-price movements compound in agents that make repeated calls, ingest long context, or generate substantial tool-use traces. At Together’s displayed rate, a simple workload using one million input tokens and one million output tokens would cost $1.85 before discounts and caching. At the earlier $0.30/$1.20 headline rates, the same illustrative mix would have cost $1.50: a 23% increase. That calculation is an example, not a prediction of a particular team’s bill.

The other listed routes make the response more nuanced than replacing the model. ModelRun and DeepInfra still show $0.30/$1.20, but selecting the cheapest route can change latency, throughput, observed tool-call errors, cache behavior, reliability, and any data-routing constraints a team operates under. A team should inspect where its requests are actually routed, compare invoice-level effective prices, and test the alternative route on representative tasks before treating a lower list price as a saving.

## What happens next
- **At least one team using this model will change or pin provider routing after comparing effective spend with latency and output quality.** (3 months, 65%). Watch for: provider-specific routing guidance or usage-share changes on OpenRouter.
- **The provider spread will remain more decision-relevant than the model’s single headline rate.** (6 months, 70%). Watch for: additional provider entries, price changes, or meaningful shifts in cache-hit and effective-rate figures.

## Limits and uncertainty
OpenRouter’s page documents the displayed provider prices but does not explain why Together’s listed rate increased or whether it will persist. The weighted averages and cache-hit figures describe aggregate observed activity, not a contractual rate for one application. A customer’s actual cost depends on input and output mix, cache behavior, discounts, routing settings, and the provider that serves a request. The performance, error-rate, uptime, and availability figures are OpenRouter measurements over recent windows, not guarantees for a given workload. The listing also does not independently establish a matching price change by Meta or by providers outside OpenRouter.

## Skill to build now
Add provider-level cost observability to an agent evaluation run this week. Log model identifier, selected provider, input tokens, output tokens, cache reads, latency, retry or failover events, and task outcome for a representative batch. Run the batch once with your current routing policy and once with an explicitly pinned lower-cost provider. Use a change only if the lower effective cost survives the quality and latency comparison; otherwise the nominal token saving is not an operational saving.

## Reader takeaway
Treat this as an OpenRouter provider-pricing signal, not evidence that every way to run Muse Glimmer 30B became more expensive. Together’s rate is higher, while two listed alternatives still show the earlier rate. The practical next step is to measure provider mix and effective spend in your own agent workload before changing routes or models.

Editor: [Andrey Reshetnikov](https://newruntime.com/owner-profile.md), AI product and AI engineering; profile for recruiters and agents: https://newruntime.com/owner-profile.json

## Continue in the graph

Next records for an agent, linked to their Markdown twins. Replace `.md` with `.json` for structured fields, including `next_reads`.

- topic hub: [Pricing](https://newruntime.com/topics/pricing.md): Explore the pricing topic hub.
- topic hub: [Inference](https://newruntime.com/topics/inference.md): Explore the inference topic hub.
- collection: [Field Notes](https://newruntime.com/posts.json): Open the latest editorial analysis.
- collection: [Sources](https://newruntime.com/sources/): See the original sources this archive cites.
- whole graph: [discovery graph](https://newruntime.com/discovery-graph.json) (typed nodes and edges) and [site index](https://newruntime.com/site-index.md)
