Grouped-Query Attention

Pick the recipe a 2026 frontier open-weight LLM block actually ships

At batch 16 and 4k context with FP8 weights, which model still fits on one 80 GB H100?

Per-token decode time grows linearly as a chat session lengthens, debug it

Expand MQA and state its KV-cache cost tradeoff

Define GQA in transformer attention

Tying K and V to one projection, what is gained and what is risked?

Spot the GQA configuration error: num_attention_heads=32, num_key_value_heads=7.

Which config token names the count of parallel attention heads in a layer?

Why does MQA underperform GQA when both share KV across heads?

Llama-2 70B uses 64 query heads, how many KV heads does it actually keep?

GQA, decode the acronym and describe its KV sharing pattern

Match GPT-2 versus Llama-2 attention design choices to their differences

Decode-phase attention is memory bandwidth bound. Explain what flips the regime.

How does DeepSeek's Multi-Latent Attention (MLA) compress the KV cache below GQA?

Match MLA, GQA and MQA to their mechanisms, cache reduction and quality posture

Predict the KV cache size in GB for given model dimensions

Which statement most accurately captures why GQA, not MQA, became the production default?

Match each attention variant to its KV-cache property and a production user

For a 64-Q-head model with GQA group size G=8, how many K/V heads exist and what is the cache reduction vs MHA?

Compare MHA, MQA, GQA, MLA, what production tradeoff are they all addressing, and which models use each?

Match each attention variant to its KV-cache memory reduction strategy.

Why is KV cache memory the bottleneck for long context LLM serving?

Compute the KV cache memory for a single request at 4096 context on Llama-2 70B (MHA) in FP16.