How much of a bottleneck is an m1 16gb Mac mini?

[worldly mountain]

For time to respond? It seems like every query takes at least 30 seconds. Using minimax m2.7 via api from official provider.

[woeful maple]

To help others find answers, you can mark your question as solved via Right click solution message -> Apps -> ✅ Mark Solution

[serene mica]

Likely not the M1 Mac mini. If you’re using MiniMax M2.7 via API, inference is happening remotely; the Mac is mostly doing gateway/session/tool orchestration. An M1/16GB can feel slow for browser/tools or lots of concurrent agents, but 30s on every simple query usually points at model/provider latency, context size, tool calls, or cold session overhead, not raw Mac CPU.

First: what OpenClaw version? Please run:

openclaw --version
openclaw status --all

Safe quick tests:

1. Try a tiny prompt: say pong only
2. Then try /fast on and repeat — for MiniMax, OpenClaw rewrites MiniMax-M2.7 to MiniMax-M2.7-highspeed on the Anthropic-compatible path. (docs.openclaw.ai)
3. If tiny prompts are still ~30s, it’s probably provider/API route latency. MiniMax officially has both MiniMax-M2.7 and MiniMax-M2.7-highspeed; highspeed is intended as the faster tier. (platform.minimax.io)

Docs: https://docs.openclaw.ai/providers/minimax

[worldly mountain]

nothing too remarkable in stats OpenClaw 2026.4.15 — updated pretty recently (last noted in the April 22 session). Everything's running on the Mac mini M1, gateway connected. /fast mode gave a pong in almost half the speed but unsure why

[serene mica]

That actually lines up: /fast on isn’t just a UI hint for MiniMax — for minimax/* it rewrites MiniMax-M2.7 → MiniMax-M2.7-highspeed, so the “pong in half the time” is probably the provider’s highspeed tier doing its job.

So I’d read this as:

• M1 Mac mini is probably not the main bottleneck for plain chat.
• MiniMax M2.7 base tier latency is probably the bottleneck if even tiny prompts take ~30s.
• If tool use/browser/file work is involved, then the Mac can add overhead, but a simple pong test isolates most of that out.

Also: 2026.4.15 is no longer super current — current npm is 2026.4.26 when I checked. I’d update first, because there have been a bunch of gateway/model/provider fixes since then, then retest.

For tuning:

/fast on

or make fast mode the default for that session/model if you’re happy with the cost/latency tradeoff.

Docs: https://docs.openclaw.ai/providers/minimax

[worldly mountain]

If I want to force open claw to enable /fast on only w one gateway device is that possible

[serene mica]

Not directly per gateway device/client, as far as the config model goes.

/fast is resolved in this order:

1. per-message /fast on
2. session override
3. per-agent agents.list[].fastModeDefault
4. per-model params.fastMode
5. off

So if multiple devices are hitting the same gateway session, /fast on is a session thing, not “only this phone/browser/device”.

Best workaround: make a separate agent for the fast path and route that channel/account/thread to it:

{

  agents: {
    list: [
      { id: "main", default: true },
      {
        id: "quick",
        fastModeDefault: true
      }
    ]
  },
  bindings: [
    {
      agentId: "quick",
      match: { channel: "telegram", peer: { kind: "direct", id: "YOUR_ID" } }
    }
  ]

}

Swap the match for whatever channel/device entry you’re trying to isolate. If you mean “one physical Gateway install only,” then yeah — just set fastModeDefault: true on that gateway’s agent config.

Docs: https://docs.openclaw.ai/tools/thinking and https://docs.openclaw.ai/channels/channel-routing

[worldly mountain]

How can I determine the current session fast mode and thinking config?

[serene mica]

Send these as standalone messages in the same chat/thread/session:

/fast

or

/fast status

Shows current effective fast mode.

/think

Shows current thinking level. Aliases:

/thinking
/t

Also useful:

/reasoning

That’s separate: it controls whether reasoning output is shown, not how much the model thinks.

And for the broad view: