#Ollama models not in model picker

1 messages · Page 1 of 1 (latest)

rapid pilot
#

A set of my local models doesn't appear in the model picker. My hunch is that the model names cause trouble. Ollama shows 5 models in the providers but I can only select 3 -- the TeichAI and bartowski ones are missing.

@native galeyour help would be appreciated!

          {
            id: "hf.co/TeichAI/Qwen3-14B-Claude-4.5-Opus-High-Reasoning-Distill-GGUF:Q4_K_M",
            name: "Qwen3-14B Claude-4.5-Opus Distill (local fallback)",
            reasoning: true,
            input: ["text"],
            cost: { input: 0, output: 0, cacheRead: 0, cacheWrite: 0 },
            contextWindow: 32768,
            maxTokens: 8192,
          },
          {
            id: "hf.co/bartowski/Qwen_Qwen3-30B-A3B-GGUF:IQ3_M",
            name: "Qwen3 30B-A3B MoE (local)",
            reasoning: true,
            input: ["text"],
            cost: { input: 0, output: 0, cacheRead: 0, cacheWrite: 0 },
            contextWindow: 32768,
            maxTokens: 8192,
          },


  agents: {
    defaults: {
      model: {
        primary: "kimi-coding/k2p5",
      },
      models: {
        // Kimi
        "kimi-coding/k2p5": { alias: "Kimi K2.5" },

        // Ollama (local)
        "ollama-local/hf.co/bartowski/Qwen_Qwen3-30B-A3B-GGUF:IQ3_M": { alias: "Qwen3 30B MoE (local)" },
        "ollama-local/qwen3:8b-q8_0": { alias: "Qwen3 8B (local)" },
        "ollama-local/hf.co/TeichAI/Qwen3-14B-Claude-4.5-Opus-High-Reasoning-Distill-GGUF:Q4_K_M": { alias: "Qwen3 14B Distill (local)" },
        "ollama-local/gemma3:12b-it-qat": { alias: "Gemma3 12B (local)" },
        "ollama-local/llama3.2:3b-instruct-q4_k_m": { alias: "Llama 3.2 3B (local)" },
      },
    },
  },
crude quarryBOT
#

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

cinder vale
#

It's because you are missing the reasoning: true flag on both the missing models. openclaw has reasoning enabled for a model, it send the system prompt using the developer role which ollama doesn't support. if you set it reasoning to false you could pick them, but you'll lose thinking mode behaviour.

#

*not missing, cause you put reasoning: true

rapid pilot
cinder vale
#

set reasoning to false and they'll show up, ollama doesn't support the reasoning: true tag, not that it doesnt have reasoning and thinking abilities, the tag does something

rapid pilot
cinder vale
#

good luck, i tried lots of local models with not much working. you find one that works probably post about it, people would want to know

rapid pilot
#

My qwen3-8b works with reasoning true - it shows up 😂

cinder vale
#

might throw an error. what it does is makes openclaw use the developer prompt, ollama doesn't support it, so even if it starts, it's going to cause issues

#

start it and message the agent internally, see what it does

rapid pilot
#

It doesn't work. Doesn't acept it.

cinder vale
#

i'm using gpt-oss:120b localy on a cluster of 3 llama.cpp nodes with rpc-servers, models needs 60GB and 14GB for context.. the only model i've ever had work, takes all my gpu's from all computers haha

rapid pilot
#

Apparently the model name throws it off

cinder vale
#

weird, that qwen3 with claude 4.5 post training sounds interestin.. never ran model before