Model Comparator

Compare up to four models using the same criteria.

2 models compared

✓ Best published value among measured models. — Data unavailable. These metrics do not establish an overall ranking.

Intelligence

Quasar 438B (Max, Based on GLM-5.2)
26.7
Agnes 2.5 Pro Alpha (based on Qwen3.5 397B A17B)
Best: 26.8

Coding

Quasar 438B (Max, Based on GLM-5.2)
Best: 61.2
Agnes 2.5 Pro Alpha (based on Qwen3.5 397B A17B)
58.8

Agentic

Quasar 438B (Max, Based on GLM-5.2)
31.4
Agnes 2.5 Pro Alpha (based on Qwen3.5 397B A17B)
—

Speed

Quasar 438B (Max, Based on GLM-5.2)
Best: 172 t/s
Agnes 2.5 Pro Alpha (based on Qwen3.5 397B A17B)
160 t/s

TTFT

Quasar 438B (Max, Based on GLM-5.2)
Best: 1.31 s
Agnes 2.5 Pro Alpha (based on Qwen3.5 397B A17B)
4.26 s

Input · $ / 1M tokens

Quasar 438B (Max, Based on GLM-5.2)
$0.60×1.3 relative to the lowest price among the compared models
Agnes 2.5 Pro Alpha (based on Qwen3.5 397B A17B)
Best: $0.45

Output · $ / 1M tokens

Quasar 438B (Max, Based on GLM-5.2)
$1.80×2 relative to the lowest price among the compared models
Agnes 2.5 Pro Alpha (based on Qwen3.5 397B A17B)
Best: $0.90

Context window

Quasar 438B (Max, Based on GLM-5.2)
1M
Agnes 2.5 Pro Alpha (based on Qwen3.5 397B A17B)
1M

Reasoning

Quasar 438B (Max, Based on GLM-5.2)
Yes
Agnes 2.5 Pro Alpha (based on Qwen3.5 397B A17B)
Yes

Open weights

Quasar 438B (Max, Based on GLM-5.2)
No
Agnes 2.5 Pro Alpha (based on Qwen3.5 397B A17B)
Yes

Cost / task

Quasar 438B (Max, Based on GLM-5.2)
$2.02
Agnes 2.5 Pro Alpha (based on Qwen3.5 397B A17B)
—

Each reasoning level remains a separate configuration.

Score and cost

Each line joins the reasoning levels of a compared model; the large point marks the configuration in the table. The top-right corner holds the best score for the lowest cost. Select another point to compare that level.

Intelligence Indexmost efficient ↗Price / 1M tokens
  • Quasar 438B (Max, Based on GLM-5.2)26.7 · $0.90
  • Agnes 2.5 Pro Alpha (based on Qwen3.5 397B A17B)26.8 · $0.563

Preferences and analytics

Rybbit: visits, performance, clicks, copies, forms and session replay (masked inputs), with your permission. Refusing does not limit the site.

Privacy

Allow navigation analytics?

With your permission, Rybbit measures visits, performance and interactions (clicks, copies, forms), including session replay, to improve BenchSift. Input values are masked in replays. Refusing does not limit the site.

Your choice is stored for 180 days. Change it at any time using “Privacy preferences” in the footer. Withdrawal reloads the page.

No choice stored: Rybbit is disabled.

Privacy