Model
Model explorer
Seed 2.2 Pro
CLOSEDByteDance · Doubao Seed family · released Sep 10, 2026
ByteDance's September 10, 2026 Seed 2.2 Pro — the second-quarterly bump to the Doubao Seed line (2.1 Pro was June 24). $0.90/$4.40 per Mtok keeps it far under every Western frontier price while the launch table pushes the vision/agentic rows ByteDance always leads with: Video-MME Pro 90.4 (a new revision it co-announces), OSWorld 3.0 74.6, MathVision 93.1. Text-side reasoning rows improve by 1.5-2.5 points across the board. As with Seed 2.1, the numbers are vendor-run but the suite is broad and consistent with ByteDance's published cadence.
ReasoningCodingVisionFunction callingTool useAgentic
2691.8
Elo · rank #19
Parameters
Undisclosed
Active params
—
Context
256K tokens
Architecture
Proprietary multimodal flagship in the Doubao Seed line; 256K-token context
License
Proprietary
Languages
—
API price (in/out)
$0.9 / $4.4
Modalities
text · vision
Benchmark results
Bar shows position within the tracked field; marker = field best
best: GPT-6 Astra · 59.3%
best: GPT-6 Astra · 94.6%
best: Kimi K3.1 · 92.4%
best: Qwen3.8-Flash-Next · 90.6%
best: this model · 88.6%
best: Claude Opus 5 · 64.7%
best: this model · 93.1%
best: this model · 91.4%
best: Claude Opus 4.7 · 85.5%
best: this model · 74.6%
best: Qwen3.7-Max · 73.6%
best: Gemini 3.8 Flash · 89.4%
best: this model · 90.4%
Run it locally
Closed weights — available via API only. No local deployment.
Input / M tok
$0.9
Output / M tok
$4.4
Doubao Seed family
Elo progression across releases
API price $0.9/$4.4 · each benchmark row carries its own source badge (see methodology)