mcpbench GitHub

client/bad-version-response

openai/gpt-5.6-sol:medium 2025-11-25 docs: none score 0% musts 1/1 · shoulds 0/1 · mays 0/0 works yes · conformant yes

Checkpoints

MUST (any failure zeroes the score — 1/1 passed)

Checkpoint Description
sent-initialize Sent an initialize request

SHOULD (the score fraction — 0/1 passed)

Checkpoint Description Detail
disconnect-on-mismatch After the server negotiated an unsupported protocol version, sent nothing further after the unsupported version response, still sent notifications/initialized

Usage & cost

Nominal cost
$0.109
Input tokens
11,858
Output tokens
1,629
Cache read
1,536
Duration
46s
Timestamp
2026-07-17 22:32:47Z
Bench version
0.3.0

Cost is nominal (public API pricing): token usage × published rates; actual marginal cost is subscription-covered.