Model

Palmyra X5

CLOSED

Writer · Palmyra X family · released Apr 28, 2025

1M-token context trained with synthetic data for ~$1M in GPU compute; reads 1M tokens in ~22s, ~300ms per function-call turn.

ReasoningCodingVisionFunction callingTool useAgentic

964.6

Elo · rank #223

Parameters

Undisclosed

Active params

—

Context

1M tokens

Architecture

Dense Transformer with adaptive/dynamic reasoning, 1M-token context (architecture undisclosed)

License

Proprietary (API-only, no released weights)

Languages

—

API price (in/out)

$0.6 / $6

Modalities

text

Benchmark results

Bar shows position within the tracked field; marker = field best

BIG-Bench HardReasoning71.0%#51

best: ERNIE 4.5 300B-A47B · 94.3%

BigCodeBenchCoding48.7%#3

best: GPT-4o mini · 57.4%

GPQA DiamondReasoning47.2%#196

best: GPT-5.6 · 94.6%

MATH-500Math71.6%#86

best: GPT-5 · 99.4%

MMLU-ProKnowledge65.0%#107

best: Claude Fable 5 · 91.5%

Run it locally

Closed weights — available via API only. No local deployment.

Input / M tok

$0.6

Output / M tok

Palmyra X family

Elo progression across releases

Resources

API price $0.6/$6 · each benchmark row carries its own source badge (see methodology)