hiboshi
V2EX  ›  问与答

有使用 opencode+omo 的吗?

  •  
  •   hiboshi · 13h 11m ago · 885 views

    昨晚刚刚开通的新的 opencode 。

    使用的是 omo 官方推荐的配置,一个简单的项目都没完成就把周限额给干完了,月限额直接 50% 了 ,是我用的姿势不对么

    {
      "$schema": "https://raw.githubusercontent.com/code-yeongyu/oh-my-openagent/dev/assets/oh-my-opencode.schema.json",
    
      "agents": {
        // Sisyphus: Kimi K2.7 is the top alternative to Claude for orchestration
        "sisyphus": {
          "model": "opencode-go/kimi-k2.7-code",
          "ultrawork": { "model": "opencode-go/kimi-k2.7-code" },
        },
    
        // Hephaestus: needs GPT. ChatGPT Plus gets you here.
        "hephaestus": { "model": "openai/gpt-5.6-sol", "variant": "medium" },
    
        // Architecture consultation: GPT or Claude Opus
        "oracle": { "model": "openai/gpt-5.5", "variant": "high" },
    
        // Prometheus keeps the same ulw-plan-backed prompt across model families
        "prometheus": { "model": "opencode-go/kimi-k2.7-code" },
    
        // Atlas also communicative — Kimi works great
        "atlas": { "model": "opencode-go/kimi-k2.7-code" },
    
        // Utility agents stay cheap
        "explore": { "model": "opencode-go/qwen3.5-plus" },
        "librarian": { "model": "opencode-go/qwen3.5-plus" },
      },
    
      "categories": {
        "visual-engineering": { "model": "opencode-go/qwen3.6-plus" },  // Qwen as Gemini alt
        "deep": { "model": "openai/gpt-5.6-terra", "variant": "xhigh" },
        "ultrabrain": { "model": "openai/gpt-5.6-sol", "variant": "xhigh" },
        "quick": { "model": "openai/gpt-5.4-mini" },
        "unspecified-low": { "model": "opencode-go/kimi-k2.7-code" },
        "unspecified-high": { "model": "opencode-go/kimi-k2.7-code" },
        "writing": { "model": "opencode-go/kimi-k2.7-code" },
      },
    
      "background_task": {
        "providerConcurrency": {
          "openai": 3,
          "opencode-go": 10,
        },
      },
    }
    
    17 replies    2026-07-13 01:58:47 +08:00
    niubilewodev
        1
    niubilewodev  
       12h 56m ago
    正常的,多 agent 就是这个德行,加上不停的切换模型,缓存率堪忧。

    我觉得 OMO 分好几个 agent 、好几种 category ,已经有点过度花活了。
    testsb
        2
    testsb  
       12h 54m ago
    一样,GO 套餐不太够用。

    做大项目,分了几个大 TASK ,每个 TASK 再分 Slice 。

    1 个 Slice 修改/新增 1000~5000+(数据、py 代码、文档),就把周限额干完了。

    后来就弃用 Go 转投 Codex Pro5x 了。
    hiboshi
        3
    hiboshi  
    OP
       12h 50m ago
    @niubilewodev 确实 搞得好复杂,是否真的有效果 目前看来我没看到结果,套餐限额就干完了。
    hiboshi
        4
    hiboshi  
    OP
       12h 49m ago
    @testsb 也准备弃掉了,话说是不是 不用 omo 是不是就直接使用 codex 省心?
    hiboshi
        5
    hiboshi  
    OP
       12h 43m ago
    @testsb 也准备弃掉了,opencode 如果不用 omo 话就没啥优势了?不如直接使用 codex ?
    testsb
        6
    testsb  
       12h 41m ago
    @hiboshi
    是的,最新的 codex 默认够用了,其实跟 omo 的工作模式差不多了。

    omo 有专门针对 Codex 的版本 https://github.com/code-yeongyu/lazycodex ,就是 omo 里写的 omo Light (Codex CLI) 版本,其实是一个 codex 的插件,调度各种 hook 、mcp 、子 agent 运行的。
    这个我也在 codex 里用过,能力是有的,做复杂编程相当于对 codex goal 模式的加强,但是加强不了太多,但是 token 消耗也比用原生 GPT 5.5 xHigh goal 大不少。

    不过最近 codex 上了 GPT 5.6 ,感觉消耗速度变大了,不用 omo 已经比以前 5.5 + omo 消耗还大了。
    hiboshi
        7
    hiboshi  
    OP
       12h 38m ago
    @testsb 好的,我用 codex 接力。
    enihcam
        8
    enihcam  
       11h 2m ago via Android
    用 omo-slim
    liuxue
        9
    liuxue  
       10h 50m ago
    放弃 omo 了,感觉 superpowers 更合适我。
    hiboshi
        10
    hiboshi  
    OP
       10h 46m ago
    @liuxue 一样 不折腾了,一样回到 superpowers 了
    nightwitch
        11
    nightwitch  
       8h 22m ago
    没什么特殊需求最好不要在不同厂商之间的模型切来切去,每个冷启动的会话都会 0 cache 命中,大幅增加 token 消耗
    imnpc
        12
    imnpc  
       7h 56m ago
    参考 OMO 的官方文档,配置下替代模型,除了核心必须得几个 GPT ,其他一律便宜模型:
    {
    "$schema": "https://raw.githubusercontent.com/code-yeongyu/oh-my-openagent/dev/assets/oh-my-opencode.schema.json",
    "agents": {
    "sisyphus": {
    "model": "opencode-go/kimi-k2.7-code",
    "fallback_models": [
    {
    "model": "opencode-go/glm-5.2"
    },
    {
    "model": "zai-coding-plan/glm-5.2"
    },
    {
    "model": "openai/gpt-5.6-sol",
    "variant": "medium"
    }
    ]
    },
    "hephaestus": {
    "model": "openai/gpt-5.6-sol",
    "variant": "medium",
    "fallback_models": [
    {
    "model": "opencode-go/deepseek-v4-pro",
    "variant": "max"
    }
    ]
    },
    "oracle": {
    "model": "opencode-go/deepseek-v4-pro",
    "variant": "max",
    "fallback_models": [
    {
    "model": "opencode-go/glm-5.2"
    },
    {
    "model": "zai-coding-plan/glm-5.2"
    },
    {
    "model": "openai/gpt-5.6-sol",
    "variant": "high"
    }
    ]
    },
    "librarian": {
    "model": "opencode-go/deepseek-v4-flash",
    "fallback_models": [
    {
    "model": "opencode-go/deepseek-v4-flash"
    },
    {
    "model": "opencode-go/qwen3.5-plus"
    },
    {
    "model": "opencode-go/minimax-m2.7"
    }
    ]
    },
    "explore": {
    "model": "opencode-go/deepseek-v4-flash",
    "fallback_models": [
    {
    "model": "opencode-go/deepseek-v4-flash"
    },
    {
    "model": "opencode-go/qwen3.5-plus"
    },
    {
    "model": "opencode-go/minimax-m2.7"
    }
    ]
    },
    "multimodal-looker": {
    "model": "opencode-go/kimi-k2.7-code",
    "fallback_models": [
    {
    "model": "opencode-go/deepseek-v4-pro"
    },
    {
    "model": "zai-coding-plan/glm-4.6v"
    }
    ]
    },
    "prometheus": {
    "model": "openai/gpt-5.6-sol",
    "variant": "high",
    "fallback_models": [
    {
    "model": "opencode-go/deepseek-v4-pro",
    "variant": "max"
    },
    {
    "model": "zai-coding-plan/glm-5.2"
    },
    {
    "model": "openai/gpt-5.6-sol",
    "variant": "high"
    }
    ]
    },
    "metis": {
    "model": "opencode-go/deepseek-v4-pro",
    "variant": "max",
    "fallback_models": [
    {
    "model": "opencode-go/glm-5.2"
    },
    {
    "model": "zai-coding-plan/glm-5.2"
    },
    {
    "model": "openai/gpt-5.6-sol",
    "variant": "high"
    }
    ]
    },
    "momus": {
    "model": "opencode-go/deepseek-v4-pro",
    "variant": "max",
    "fallback_models": [
    {
    "model": "opencode-go/glm-5.2"
    },
    {
    "model": "zai-coding-plan/glm-5.2"
    },
    {
    "model": "openai/gpt-5.6-sol",
    "variant": "xhigh"
    }
    ]
    },
    "atlas": {
    "model": "opencode-go/deepseek-v4-pro",
    "variant": "max",
    "fallback_models": [
    {
    "model": "opencode-go/kimi-k2.7-code"
    },
    {
    "model": "openai/gpt-5.6-sol",
    "variant": "medium"
    },
    {
    "model": "opencode-go/minimax-m2.7"
    }
    ]
    },
    "sisyphus-junior": {
    "model": "opencode-go/deepseek-v4-pro",
    "variant": "max",
    "fallback_models": [
    {
    "model": "opencode-go/kimi-k2.7-code"
    },
    {
    "model": "openai/gpt-5.6-sol",
    "variant": "medium"
    },
    {
    "model": "opencode-go/minimax-m2.7"
    }
    ]
    }
    },
    "categories": {
    "visual-engineering": {
    "model": "zai-coding-plan/glm-5.2",
    "fallback_models": [
    {
    "model": "opencode-go/glm-5.2"
    }
    ]
    },
    "ultrabrain": {
    "model": "opencode-go/deepseek-v4-pro",
    "variant": "max",
    "fallback_models": [
    {
    "model": "openai/gpt-5.6-sol",
    "variant": "xhigh"
    },
    {
    "model": "opencode-go/glm-5.2"
    },
    {
    "model": "zai-coding-plan/glm-5.2"
    }
    ]
    },
    "deep": {
    "model": "opencode-go/deepseek-v4-pro",
    "variant": "max",
    "fallback_models": [
    {
    "model": "opencode-go/kimi-k2.7-code"
    },
    {
    "model": "opencode-go/glm-5.2"
    },
    {
    "model": "zai-coding-plan/glm-5.2"
    },
    {
    "model": "openai/gpt-5.6-sol",
    "variant": "medium"
    }
    ]
    },
    "artistry": {
    "model": "opencode-go/deepseek-v4-pro",
    "variant": "max",
    "fallback_models": [
    {
    "model": "opencode-go/kimi-k2.7-code"
    },
    {
    "model": "opencode-go/glm-5.2"
    },
    {
    "model": "openai/gpt-5.6-sol"
    }
    ]
    },
    "quick": {
    "model": "opencode-go/deepseek-v4-flash",
    "fallback_models": [
    {
    "model": "opencode-go/deepseek-v4-flash"
    },
    {
    "model": "opencode-go/minimax-m2.7"
    },
    {
    "model": "openai/gpt-5.4-mini"
    }
    ]
    },
    "unspecified-low": {
    "model": "opencode-go/deepseek-v4-flash",
    "fallback_models": [
    {
    "model": "opencode-go/deepseek-v4-flash"
    },
    {
    "model": "opencode-go/kimi-k2.7-code"
    },
    {
    "model": "opencode-go/minimax-m2.7"
    },
    {
    "model": "openai/gpt-5.3-codex"
    }
    ]
    },
    "unspecified-high": {
    "model": "opencode-go/deepseek-v4-pro",
    "variant": "max",
    "fallback_models": [
    {
    "model": "opencode-go/glm-5.2"
    },
    {
    "model": "zai-coding-plan/glm-5.2"
    },
    {
    "model": "opencode-go/kimi-k2.7-code"
    },
    {
    "model": "opencode-go/minimax-m2.7"
    },
    {
    "model": "openai/gpt-5.6-sol",
    "variant": "high"
    }
    ]
    },
    "writing": {
    "model": "opencode-go/kimi-k2.7-code",
    "fallback_models": [
    {
    "model": "opencode-go/minimax-m2.7"
    }
    ]
    }
    }
    }
    hiboshi
        13
    hiboshi  
    OP
       7h 16m ago
    @imnpc sisyphus 用的 opencode-go/kimi-k2.7-code 就不便宜,输出四刀了,我一个接口需求没做完就没额度了。至少在 opencode-go 的套餐上是不够用的。
    hiboshi
        14
    hiboshi  
    OP
       7h 14m ago
    @nightwitch 确实,没啥特殊的需求,我只是不想被 agent 厂商绑架没有选择 cc 、codex,选择了 opencode ,但是纯 op 不是很好用就用了 omo 了
    maolon
        15
    maolon  
       6h 24m ago
    omo 是特别重的一个 harness ,只有在大型项目下才好用(减少漂移),
    另外 opencode go 套餐本来就给的少,那更不经用了。然后你还选了配额消耗大的模型做 sisyphus ( kimi k2.7 code ),如果要用 go 做主力考虑小米的模型和 deepseek 的,这两是量给的最慷慨能力又算在线的
    frank553000
        16
    frank553000  
       6h 8m ago
    我让网页版的 openai 和 claudecode 帮我生成了一个,优先使用 deepseek flash 和 mino 套餐,用的时候还好,有些复杂功能会自动切换到套餐少的部分
    dreampuf
        17
    dreampuf  
       3h 15m ago
    omo 不好用,或者说整体这种企图靠 prompt 来约束的方案就不靠谱。更多是靠 base model 本身的能力,以及适当的问题拆分。

    花费了几百 B 得到的结论。
    About   ·   Help   ·   Advertise   ·   Blog   ·   API   ·   FAQ   ·   Solana   ·   846 Online   Highest 6679   ·     Select Language
    创意工作者们的社区
    World is powered by solitude
    VERSION: 3.9.8.5 · 50ms · UTC 21:14 · PVG 05:14 · LAX 14:14 · JFK 17:14
    ♥ Do have faith in what you're doing.