Ember-1

Ember-1 is a reasoning model from Fireworks in the Fireworks Ember family. It has 8 scored results in 3 categories. It has no index.

  • Proprietary
  • Reasoning
  • Input: text · image
  • Released 2026-09
partialShows if the model has enough results for an index.
IndexOverall score out of 100.UnrankedMedian 50.7
CoverageShare of the index weight with results.50%
SpeedOutput tokens per second.48/sMedian 62/s
First tokenSeconds to the first output token.3.1 sMedian 1.6 s
Input / 1MUS dollars per 1M input tokens.$3Median $0.6
Output / 1MUS dollars per 1M output tokens.$15Median $2.5
ContextMaximum tokens in one request.1.05MMedian 262K
EloLMArena rating and rank.N/AMedian 1427

Details

From OpenRouter
API IDThe model ID on OpenRouter.
fireworks/ember-1
Max outputMaximum output tokens per request.
944K
Knowledge cutoffLast date of training data.
N/A
ToolsTool calls on OpenRouter.
Yes
JSON outputOutput follows a given JSON schema.
Yes
Cache read / 1MUS dollars per 1M cached tokens.
$0.3
WeightsLink to the published weights.
N/A
RetiresDate OpenRouter removes the model.
N/A

Capabilities

Score per category, out of 100.
Out of 100
AgenticMulti-step tasks with tools.3 counted71.5—
CodingCode writing and repair.4 counted67.7—
ReasoningLogic problems and puzzles.None countedN/A—
MultimodalTasks with images and text.None countedN/A—
KnowledgeFacts and expert knowledge.None countedN/A—
MultilingualTasks in many languages.None countedN/A—
InstructionTasks with strict rules in the prompt.None countedN/A—
MathMath problems.1 counted70.3—

Results

8 counted
BenchmarkThe test name.ResultThe score from the source.PlacePlace among results on this benchmark.
DeepSWEAgentic75.2%3 / 41
Vibe Code Bench v1.1Coding83.9%15 / 107
ProofBench v1.1Math61.0%20 / 49
Terminal-Bench 4.0Agentic19.7%21 / 44
Code MigrationCoding32.5%35 / 73
IOICoding46.3%27 / 41
Tax Agent BenchFinance68.0%12 / 66
Legal Research BenchLegal44.7%17 / 73
Public Benefits Bench v1.1Public benefits66.2%17 / 47
EMBFinance63.4%26 / 70
Finance Agent (v2)Finance51.9%33 / 74
CyberBench v1.1Other69.5%20 / 44
Terminal-Bench ScienceScience5.7%18 / 38
BioMysteryBenchScience67.0%16 / 24
MysteryMechanismScience22.5%16 / 24

Sources

BenchLM benchmark aggregationCC BY-NC 4.0 · Data from BenchLM.aiOpenRouter, collected directlyNo licence statedVals AI, collected directlyNo licence stated. Read from the public leaderboard and credited to Vals AI