Closed SourceWriterReleased 2026-08

Palmyra X6

Frontier-level performance for marketing and revenue teams with dramatically lower costs

Context window

1000000

Input / 1M tokens

$2.00

Output / 1M tokens

$8.00

Provider

Writer

$2.00 per 1M input tokens and $8.00 per 1M output tokens on Writer's API (about $3.50 blended at a 3:1 mix, per Writer), with an 8,192-token output limit. Writer is deprecating Palmyra X4 and X5 on December 14, 2026 and names X6 as the migration path. · Data verified 2026-11-08

Palmyra X6 is a 744-billion-parameter mixture-of-experts model with roughly 40 billion active parameters per token, built as a post-training variation on GLM-5.2's architecture. Engineered for cost efficiency at scale, it delivers frontier-competitive performance while keeping token consumption costs low. The model is optimized for agentic work and long-running tasks across high-volume enterprise workflows.

Capability index

Relative estimates (0-100) to place this model against its peers, grounded in published benchmarks.

Coding
70
Reasoning
80
Math
70
Multimodal
0
Long context
70
Speed
80
Cost efficiency
90

How to access it

Writer's API (AI Studio, model ID palmyra-x6) and the WRITER Agent platform. Writer's model docs list no Amazon Bedrock ID for X6, unlike the older X4 and X5.

Strengths

  • ✓Paired with WRITER Agent: 52% lower cost, 48% faster and 10% higher quality on average than the previous generation (Writer-reported)
  • ✓Per-task cost and latency roughly halved versus the previous generation, with no measured quality regressions (Writer-reported)
  • ✓Can run unattended on a single goal for up to 8 hours
  • ✓1M-token context window

Best for developers who...

High-volume marketing and sales AI agentsCost-sensitive enterprise deployments at scaleLong-running agentic workflowsContent generation and personalization at scale

When to choose it (and when not to)

Reach for Palmyra X6 when...

  • →Enterprise teams needing to control token spending while scaling AI agents
  • →Organizations running agents across thousands of employees
  • →Cost-sensitive workloads requiring frontier-level reasoning

Look elsewhere if...

  • ✕You need open weights (Writer does not publish X6 weights; the base model GLM-5.2 is open)
  • ✕You need the model on Amazon Bedrock or Azure (Writer lists no Bedrock ID for X6)
  • ✕You need responses longer than 8,192 tokens

How to use it

  • ›Not specified in official documentation

Quickstart

Not specified
Not specified

Access through Writer's AI Studio, Agent Builder, or API

API model id: palmyra-x6

Benchmarks

BenchmarkScoreNotes
Internal Writer Enterprise Evaluation (9 capabilities)0.87/1.00Covers grounding, retrieval, tool use, content generation, sub-agent delegation, brand voice; outperformed Claude Opus 4.8 (0.86), Claude Sonnet 4.6 (0.85), GPT-5.5 (0.80), Gemini 3.1 (0.77)
FORTRESS Adversarial Safety Benchmark8.6 points higher than GLM-5.2 baseWith negligible cost to benign helpfulness

Source: Writer Official Blog

Compare Palmyra X6

Compare Palmyra X6 with any other model

Build a comparison →
All model comparisons →

Learn the concepts