DROPGoogle launches Gemini 3.8 Flash TTS and Flash-Lite TTS///DROPGemini Omni brings 1080p generation to Google Vids///BREAKClaude Opus 5.5 is here: 1M context, $4 input / $20 output///BREAKOpenAI adds GPT-6 Sol and Luna to the GPT-6 family///PAPEROpus 5.5's system card documents gains—and evaluation limits///BREAKGrok 4.7 launches at $2 input / $6 output per million tokens///DROPGrok Voice Transcribe 2.0 targets difficult real-world audio///TRENDGoogle steers new Gemini API projects to 3.8 Flash or 3.5 Flash-Lite///DROPGoogle launches Gemini 3.8 Flash TTS and Flash-Lite TTS///DROPGemini Omni brings 1080p generation to Google Vids///BREAKClaude Opus 5.5 is here: 1M context, $4 input / $20 output///BREAKOpenAI adds GPT-6 Sol and Luna to the GPT-6 family///PAPEROpus 5.5's system card documents gains—and evaluation limits///BREAKGrok 4.7 launches at $2 input / $6 output per million tokens///DROPGrok Voice Transcribe 2.0 targets difficult real-world audio///TRENDGoogle steers new Gemini API projects to 3.8 Flash or 3.5 Flash-Lite///
DARC.DEV

[ LAB — LIVE · no model calls ]

Local Memory Planner

Estimate raw weight storage from parameter count and precision, add an explicit planning allowance and compare it with available memory. See both decimal GB and binary GiB. The tool deliberately avoids promising that a model fits or will run quickly.

All lab tools

Interactive workbench / no model calls

Use total checkpoint parameters, not MoE active parameters. Available memory is in GiB; results distinguish decimal GB from GiB.

Inputs are sent to this application's server for deterministic calculation and are not stored. Do not paste secrets or raw customer content.

How to use it

  1. 01

    Enter total checkpoint parameters, not only active parameters for an MoE model.

  2. 02

    Choose nominal bits per weight and a separate percentage allowance for planning.

  3. 03

    Compare with available memory and measure the real runtime under your target context and concurrency.

Sources and methodology