[BUG]

[gwynnbleiidd]

🐛 What happened?

🔄 Steps to reproduce

1. Type message
2. Click send

✅ What did you expect?

Normal answer

🖥️ Environment

• Extension version: 0.4.2
• VS Code version: 1.130.0
• Provider: OpenCode Go
• Model: mimo-v2.5, deepseek-v4-flash and other

📸 Logs or screenshots (optional)

# OpenCode Go Diagnostics

## Recent Requests

### 1. hy3

- time: 2026-07-24T08:43:18.943Z
- endpoint: chat-completions
- initiator: github.copilot-chat
- metadataSource: models.dev
- status: 200
- durationMs: 97122
- ttfbMs: 10556
- totalBytes: 0
- totalEvents: 0
- tokens: prompt=n/a, completion=n/a, total=n/a, cached=n/a
- cacheHitRatio: n/a
- finishReason: n/a
- requestId: req-8ca3ccb7
- sessionId: vscode-8249604f
- url: https://opencode.ai/zen/go/v1/chat/completions
- error: request cancelled

### 2. hy3

- time: 2026-07-24T08:41:35.820Z
- endpoint: chat-completions
- initiator: github.copilot-chat
- metadataSource: models.dev
- status: 200
- durationMs: 100998
- ttfbMs: 10701
- totalBytes: 0
- totalEvents: 0
- tokens: prompt=n/a, completion=n/a, total=n/a, cached=n/a
- cacheHitRatio: n/a
- finishReason: n/a
- requestId: req-b0345245
- sessionId: vscode-18482757
- url: https://opencode.ai/zen/go/v1/chat/completions
- error: request cancelled

### 3. mimo-v2.5

- time: 2026-07-24T08:38:07.797Z
- endpoint: chat-completions
- initiator: github.copilot-chat
- metadataSource: models.dev
- status: 200
- durationMs: 4029
- ttfbMs: 3974
- totalBytes: 976
- totalEvents: 5
- tokens: prompt=754773, completion=1, total=754774, cached=754752
- cacheHitRatio: 100.0%
- finishReason: length
- requestId: req-bc031363
- sessionId: vscode-086ceb80
- url: https://opencode.ai/zen/go/v1/chat/completions

### 4. mimo-v2.5

- time: 2026-07-24T08:38:01.996Z
- endpoint: chat-completions
- initiator: github.copilot-chat
- metadataSource: models.dev
- status: 200
- durationMs: 22956
- ttfbMs: 6360
- totalBytes: 1032
- totalEvents: 9
- tokens: prompt=754773, completion=1, total=754774, cached=754752
- cacheHitRatio: 100.0%
- finishReason: length
- requestId: req-22db5a2f
- sessionId: vscode-086ceb80
- url: https://opencode.ai/zen/go/v1/chat/completions

### 5. mimo-v2.5

- time: 2026-07-24T08:37:37.273Z
- endpoint: chat-completions
- initiator: github.copilot-chat
- metadataSource: models.dev
- status: 200
- durationMs: 13420
- ttfbMs: 6200
- totalBytes: 1004
- totalEvents: 7
- tokens: prompt=754773, completion=1, total=754774, cached=754752
- cacheHitRatio: 100.0%
- finishReason: length
- requestId: req-03b07440
- sessionId: vscode-086ceb80
- url: https://opencode.ai/zen/go/v1/chat/completions

### 6. mimo-v2.5

- time: 2026-07-24T08:37:21.808Z
- endpoint: chat-completions
- initiator: github.copilot-chat
- metadataSource: models.dev
- status: 200
- durationMs: 29341
- ttfbMs: 6995
- totalBytes: 1046
- totalEvents: 10
- tokens: prompt=754773, completion=1, total=754774, cached=606208
- cacheHitRatio: 80.3%
- finishReason: length
- requestId: req-e54f09dc
- sessionId: vscode-086ceb80
- url: https://opencode.ai/zen/go/v1/chat/completions

### 7. mimo-v2.5

- time: 2026-07-24T08:32:16.046Z
- endpoint: chat-completions
- initiator: github.copilot-chat
- metadataSource: models.dev
- status: 200
- durationMs: 7249
- ttfbMs: 6222
- totalBytes: 990
- totalEvents: 6
- tokens: prompt=754772, completion=1, total=754773, cached=754752
- cacheHitRatio: 100.0%
- finishReason: length
- requestId: req-b79ce429
- sessionId: vscode-086ceb80
- url: https://opencode.ai/zen/go/v1/chat/completions

### 8. mimo-v2.5

- time: 2026-07-24T08:32:07.302Z
- endpoint: chat-completions
- initiator: github.copilot-chat
- metadataSource: models.dev
- status: 200
- durationMs: 4853
- ttfbMs: 4785
- totalBytes: 976
- totalEvents: 5
- tokens: prompt=754772, completion=1, total=754773, cached=754752
- cacheHitRatio: 100.0%
- finishReason: length
- requestId: req-8b604721
- sessionId: vscode-086ceb80
- url: https://opencode.ai/zen/go/v1/chat/completions

### 9. mimo-v2.5

- time: 2026-07-24T08:32:00.908Z
- endpoint: chat-completions
- initiator: github.copilot-chat
- metadataSource: models.dev
- status: 200
- durationMs: 5405
- ttfbMs: 5362
- totalBytes: 976
- totalEvents: 5
- tokens: prompt=754772, completion=1, total=754773, cached=754752
- cacheHitRatio: 100.0%
- finishReason: length
- requestId: req-ac25e425
- sessionId: vscode-086ceb80
- url: https://opencode.ai/zen/go/v1/chat/completions

### 10. mimo-v2.5

- time: 2026-07-24T08:31:53.522Z
- endpoint: chat-completions
- initiator: github.copilot-chat
- metadataSource: models.dev
- status: 200
- durationMs: 49842
- ttfbMs: 6710
- totalBytes: 1102
- totalEvents: 14
- tokens: prompt=754772, completion=1, total=754773, cached=606208
- cacheHitRatio: 80.3%
- finishReason: length
- requestId: req-e337bbdb
- sessionId: vscode-086ceb80
- url: https://opencode.ai/zen/go/v1/chat/completions

### 11. mimo-v2.5

- time: 2026-07-24T08:21:26.570Z
- endpoint: chat-completions
- initiator: github.copilot-chat
- metadataSource: models.dev
- status: 200
- durationMs: 8589
- ttfbMs: 6949
- totalBytes: 990
- totalEvents: 6
- tokens: prompt=754737, completion=1, total=754738, cached=754688
- cacheHitRatio: 100.0%
- finishReason: length
- requestId: req-9a033be5
- sessionId: vscode-086ceb80
- url: https://opencode.ai/zen/go/v1/chat/completions

### 12. mimo-v2.5

- time: 2026-07-24T08:21:16.077Z
- endpoint: chat-completions
- initiator: github.copilot-chat
- metadataSource: models.dev
- status: 200
- durationMs: 25166
- ttfbMs: 8703
- totalBytes: 1362
- totalEvents: 10
- tokens: prompt=754737, completion=1, total=754738, cached=754688
- cacheHitRatio: 100.0%
- finishReason: length
- requestId: req-d833b2ac
- sessionId: vscode-086ceb80
- url: https://opencode.ai/zen/go/v1/chat/completions

### 13. mimo-v2.5

- time: 2026-07-24T08:20:49.344Z
- endpoint: chat-completions
- initiator: github.copilot-chat
- metadataSource: models.dev
- status: 200
- durationMs: 7775
- ttfbMs: 7575
- totalBytes: 1320
- totalEvents: 7
- tokens: prompt=754737, completion=1, total=754738, cached=754688
- cacheHitRatio: 100.0%
- finishReason: length
- requestId: req-9eb5d902
- sessionId: vscode-086ceb80
- url: https://opencode.ai/zen/go/v1/chat/completions

### 14. mimo-v2.5

- time: 2026-07-24T08:20:39.559Z
- endpoint: chat-completions
- initiator: github.copilot-chat
- metadataSource: models.dev
- status: 200
- durationMs: 81189
- ttfbMs: 8622
- totalBytes: 1518
- totalEvents: 21
- tokens: prompt=754737, completion=1, total=754738, cached=16384
- cacheHitRatio: 2.2%
- finishReason: length
- requestId: req-207ac488
- sessionId: vscode-086ceb80
- url: https://opencode.ai/zen/go/v1/chat/completions

### 15. deepseek-v4-flash

- time: 2026-07-24T08:18:51.696Z
- endpoint: chat-completions
- initiator: github.copilot-chat
- metadataSource: models.dev
- status: cancelled
- durationMs: 2880
- ttfbMs: n/a
- totalBytes: 0
- totalEvents: 0
- tokens: prompt=n/a, completion=n/a, total=n/a, cached=n/a
- cacheHitRatio: n/a
- finishReason: n/a
- requestId: req-4594af6e
- sessionId: vscode-5404d53e
- url: https://opencode.ai/zen/go/v1/chat/completions
- error: request cancelled

### 16. deepseek-v4-flash

- time: 2026-07-24T08:18:28.974Z
- endpoint: chat-completions
- initiator: github.copilot-chat
- metadataSource: models.dev
- status: cancelled
- durationMs: 1296
- ttfbMs: n/a
- totalBytes: 0
- totalEvents: 0
- tokens: prompt=n/a, completion=n/a, total=n/a, cached=n/a
- cacheHitRatio: n/a
- finishReason: n/a
- requestId: req-3659f90d
- sessionId: vscode-5404d53e
- url: https://opencode.ai/zen/go/v1/chat/completions
- error: request cancelled

### 17. deepseek-v4-flash

- time: 2026-07-24T08:18:25.773Z
- endpoint: chat-completions
- initiator: github.copilot-chat
- metadataSource: models.dev
- status: 200
- durationMs: 5918
- ttfbMs: 4811
- totalBytes: 1400
- totalEvents: 6
- tokens: prompt=754122, completion=1, total=754123, cached=754048
- cacheHitRatio: 100.0%
- finishReason: length
- requestId: req-d056a3c4
- sessionId: vscode-5404d53e
- url: https://opencode.ai/zen/go/v1/chat/completions

### 18. deepseek-v4-flash

- time: 2026-07-24T08:18:00.915Z
- endpoint: chat-completions
- initiator: github.copilot-chat
- metadataSource: models.dev
- status: 200
- durationMs: 5268
- ttfbMs: 4122
- totalBytes: 1400
- totalEvents: 6
- tokens: prompt=754122, completion=1, total=754123, cached=754048
- cacheHitRatio: 100.0%
- finishReason: length
- requestId: req-0e7ef14b
- sessionId: vscode-5404d53e
- url: https://opencode.ai/zen/go/v1/chat/completions

### 19. deepseek-v4-flash

- time: 2026-07-24T08:17:54.072Z
- endpoint: chat-completions
- initiator: github.copilot-chat
- metadataSource: models.dev
- status: 200
- durationMs: 4117
- ttfbMs: 3302
- totalBytes: 1400
- totalEvents: 6
- tokens: prompt=754122, completion=1, total=754123, cached=754048
- cacheHitRatio: 100.0%
- finishReason: length
- requestId: req-5dd59e04
- sessionId: vscode-5404d53e
- url: https://opencode.ai/zen/go/v1/chat/completions

### 20. deepseek-v4-flash

- time: 2026-07-24T08:17:48.399Z
- endpoint: chat-completions
- initiator: github.copilot-chat
- metadataSource: models.dev
- status: 200
- durationMs: 5440
- ttfbMs: 4626
- totalBytes: 1400
- totalEvents: 6
- tokens: prompt=754122, completion=1, total=754123, cached=754048
- cacheHitRatio: 100.0%
- finishReason: length
- requestId: req-ae8d7dec
- sessionId: vscode-5404d53e
- url: https://opencode.ai/zen/go/v1/chat/completions

### 21. deepseek-v4-flash

- time: 2026-07-24T08:17:41.269Z
- endpoint: chat-completions
- initiator: github.copilot-chat
- metadataSource: models.dev
- status: 200
- durationMs: 6594
- ttfbMs: 4475
- totalBytes: 1408
- totalEvents: 6
- tokens: prompt=754122, completion=1, total=754123, cached=643712
- cacheHitRatio: 85.4%
- finishReason: length
- requestId: req-e2783d57
- sessionId: vscode-5404d53e
- url: https://opencode.ai/zen/go/v1/chat/completions

### 22. deepseek-v4-flash

- time: 2026-07-24T08:16:33.873Z
- endpoint: chat-completions
- initiator: github.copilot-chat
- metadataSource: models.dev
- status: 200
- durationMs: 4518
- ttfbMs: 3944
- totalBytes: 1400
- totalEvents: 6
- tokens: prompt=739105, completion=1, total=739106, cached=739072
- cacheHitRatio: 100.0%
- finishReason: length
- requestId: req-a3857d3f
- sessionId: vscode-5404d53e
- url: https://opencode.ai/zen/go/v1/chat/completions

### 23. deepseek-v4-flash

- time: 2026-07-24T08:16:27.845Z
- endpoint: chat-completions
- initiator: github.copilot-chat
- metadataSource: models.dev
- status: 200
- durationMs: 4680
- ttfbMs: 3668
- totalBytes: 1400
- totalEvents: 6
- tokens: prompt=739105, completion=1, total=739106, cached=739072
- cacheHitRatio: 100.0%
- finishReason: length
- requestId: req-e22fe887
- sessionId: vscode-5404d53e
- url: https://opencode.ai/zen/go/v1/chat/completions

### 24. deepseek-v4-flash

- time: 2026-07-24T08:16:21.650Z
- endpoint: chat-completions
- initiator: github.copilot-chat
- metadataSource: models.dev
- status: 200
- durationMs: 4543
- ttfbMs: 3413
- totalBytes: 1406
- totalEvents: 6
- tokens: prompt=739105, completion=1, total=739106, cached=720896
- cacheHitRatio: 97.5%
- finishReason: length
- requestId: req-2927ba46
- sessionId: vscode-5404d53e
- url: https://opencode.ai/zen/go/v1/chat/completions

### 25. deepseek-v4-flash

- time: 2026-07-24T08:16:15.458Z
- endpoint: chat-completions
- initiator: github.copilot-chat
- metadataSource: models.dev
- status: 200
- durationMs: 6668
- ttfbMs: 4168
- totalBytes: 1406
- totalEvents: 6
- tokens: prompt=739105, completion=1, total=739106, cached=643712
- cacheHitRatio: 87.1%
- finishReason: length
- requestId: req-09f50218
- sessionId: vscode-5404d53e
- url: https://opencode.ai/zen/go/v1/chat/completions

## Models

Models visible through vscode.lm.selectChatModels({ vendor: "opencodego" }): 23

- minimax-m3
  rawModelId: minimax-m3
  name: OpenCode Go / Minimax M3
  family: opencodego-minimax-m3-session-2026-05-21-b
  vendor: opencodego
  version: 1.2.0-session-2026-05-21-b-1000000-131072
  maxInputTokens: 991808
  advertisedMaxOutputTokens: 8192
  advertisedContextWindow: 1000000
  apiMaxOutputTokens: 131072
  metadataSource: models.dev
  supportsVision: true
  status: active
  thinkingFamily: minimax
  configurationSchema: null
- minimax-m2.7
  rawModelId: minimax-m2.7
  name: OpenCode Go / Minimax M2.7
  family: opencodego-minimax-m2.7-session-2026-05-21-b
  vendor: opencodego
  version: 1.2.0-session-2026-05-21-b-204800-40896
  maxInputTokens: 196608
  advertisedMaxOutputTokens: 8192
  advertisedContextWindow: 204800
  apiMaxOutputTokens: 40896
  metadataSource: models.dev
  supportsVision: false
  status: active
  thinkingFamily: minimax
  configurationSchema: null
- minimax-m2.5
  rawModelId: minimax-m2.5
  name: OpenCode Go / Minimax M2.5
  family: opencodego-minimax-m2.5-session-2026-05-21-b
  vendor: opencodego
  version: 1.2.0-session-2026-05-21-b-204800-40896
  maxInputTokens: 196608
  advertisedMaxOutputTokens: 8192
  advertisedContextWindow: 204800
  apiMaxOutputTokens: 40896
  metadataSource: models.dev
  supportsVision: false
  status: deprecated
  thinkingFamily: minimax
  configurationSchema: null
- kimi-k3
  rawModelId: kimi-k3
  name: OpenCode Go / Kimi K3
  family: opencodego-kimi-k3-session-2026-05-21-b
  vendor: opencodego
  version: 1.2.0-session-2026-05-21-b-1048576-131072
  maxInputTokens: 1040384
  advertisedMaxOutputTokens: 8192
  advertisedContextWindow: 1048576
  apiMaxOutputTokens: 131072
  metadataSource: models.dev
  supportsVision: true
  status: active
  thinkingFamily: kimi
  configurationSchema: null
- kimi-k2.7-code
  rawModelId: kimi-k2.7-code
  name: OpenCode Go / Kimi K2.7 Code
  family: opencodego-kimi-k2.7-code-session-2026-05-21-b
  vendor: opencodego
  version: 1.2.0-session-2026-05-21-b-262144-52365
  maxInputTokens: 253952
  advertisedMaxOutputTokens: 8192
  advertisedContextWindow: 262144
  apiMaxOutputTokens: 52365
  metadataSource: models.dev
  supportsVision: true
  status: active
  thinkingFamily: kimi
  configurationSchema: null
- kimi-k2.6
  rawModelId: kimi-k2.6
  name: OpenCode Go / Kimi K2.6
  family: opencodego-kimi-k2.6-session-2026-05-21-b
  vendor: opencodego
  version: 1.2.0-session-2026-05-21-b-262144-52365
  maxInputTokens: 253952
  advertisedMaxOutputTokens: 8192
  advertisedContextWindow: 262144
  apiMaxOutputTokens: 52365
  metadataSource: models.dev
  supportsVision: true
  status: active
  thinkingFamily: kimi
  configurationSchema: null
- kimi-k2.5
  rawModelId: kimi-k2.5
  name: OpenCode Go / Kimi K2.5
  family: opencodego-kimi-k2.5-session-2026-05-21-b
  vendor: opencodego
  version: 1.2.0-session-2026-05-21-b-262144-52365
  maxInputTokens: 253952
  advertisedMaxOutputTokens: 8192
  advertisedContextWindow: 262144
  apiMaxOutputTokens: 52365
  metadataSource: models.dev
  supportsVision: true
  status: deprecated
  thinkingFamily: kimi
  configurationSchema: null
- glm-5.2
  rawModelId: glm-5.2
  name: OpenCode Go / Glm 5.2
  family: opencodego-glm-5.2-session-2026-05-21-b
  vendor: opencodego
  version: 1.2.0-session-2026-05-21-b-1000000-131072
  maxInputTokens: 991808
  advertisedMaxOutputTokens: 8192
  advertisedContextWindow: 1000000
  apiMaxOutputTokens: 131072
  metadataSource: models.dev
  supportsVision: false
  status: active
  thinkingFamily: glm
  configurationSchema: null
- glm-5.1
  rawModelId: glm-5.1
  name: OpenCode Go / Glm 5.1
  family: opencodego-glm-5.1-session-2026-05-21-b
  vendor: opencodego
  version: 1.2.0-session-2026-05-21-b-202752-32768
  maxInputTokens: 194560
  advertisedMaxOutputTokens: 8192
  advertisedContextWindow: 202752
  apiMaxOutputTokens: 32768
  metadataSource: models.dev
  supportsVision: false
  status: active
  thinkingFamily: glm
  configurationSchema: null
- glm-5
  rawModelId: glm-5
  name: OpenCode Go / Glm 5
  family: opencodego-glm-5-session-2026-05-21-b
  vendor: opencodego
  version: 1.2.0-session-2026-05-21-b-202752-32768
  maxInputTokens: 194560
  advertisedMaxOutputTokens: 8192
  advertisedContextWindow: 202752
  apiMaxOutputTokens: 32768
  metadataSource: models.dev
  supportsVision: false
  status: deprecated
  thinkingFamily: glm
  configurationSchema: null
- deepseek-v4-pro
  rawModelId: deepseek-v4-pro
  name: OpenCode Go / Deepseek V4 Pro
  family: opencodego-deepseek-v4-pro-session-2026-05-21-b
  vendor: opencodego
  version: 1.2.0-session-2026-05-21-b-1000000-199936
  maxInputTokens: 991808
  advertisedMaxOutputTokens: 8192
  advertisedContextWindow: 1000000
  apiMaxOutputTokens: 199936
  metadataSource: models.dev
  supportsVision: false
  status: active
  thinkingFamily: deepseek
  configurationSchema: null
- deepseek-v4-flash
  rawModelId: deepseek-v4-flash
  name: OpenCode Go / Deepseek V4 Flash
  family: opencodego-deepseek-v4-flash-session-2026-05-21-b
  vendor: opencodego
  version: 1.2.0-session-2026-05-21-b-1000000-199936
  maxInputTokens: 991808
  advertisedMaxOutputTokens: 8192
  advertisedContextWindow: 1000000
  apiMaxOutputTokens: 199936
  metadataSource: models.dev
  supportsVision: false
  status: active
  thinkingFamily: deepseek
  configurationSchema: null
- qwen3.7-max
  rawModelId: qwen3.7-max
  name: OpenCode Go / Qwen3.7 Max
  family: opencodego-qwen3.7-max-session-2026-05-21-b
  vendor: opencodego
  version: 1.2.0-session-2026-05-21-b-1000000-65536
  maxInputTokens: 991808
  advertisedMaxOutputTokens: 8192
  advertisedContextWindow: 1000000
  apiMaxOutputTokens: 65536
  metadataSource: models.dev
  supportsVision: false
  status: active
  thinkingFamily: qwen
  configurationSchema: null
- qwen3.7-plus
  rawModelId: qwen3.7-plus
  name: OpenCode Go / Qwen3.7 Plus
  family: opencodego-qwen3.7-plus-session-2026-05-21-b
  vendor: opencodego
  version: 1.2.0-session-2026-05-21-b-1000000-65536
  maxInputTokens: 991808
  advertisedMaxOutputTokens: 8192
  advertisedContextWindow: 1000000
  apiMaxOutputTokens: 65536
  metadataSource: models.dev
  supportsVision: true
  status: active
  thinkingFamily: qwen
  configurationSchema: null
- qwen3.6-plus
  rawModelId: qwen3.6-plus
  name: OpenCode Go / Qwen3.6 Plus
  family: opencodego-qwen3.6-plus-session-2026-05-21-b
  vendor: opencodego
  version: 1.2.0-session-2026-05-21-b-1000000-65536
  maxInputTokens: 991808
  advertisedMaxOutputTokens: 8192
  advertisedContextWindow: 1000000
  apiMaxOutputTokens: 65536
  metadataSource: models.dev
  supportsVision: true
  status: active
  thinkingFamily: qwen
  configurationSchema: null
- qwen3.5-plus
  rawModelId: qwen3.5-plus
  name: OpenCode Go / Qwen3.5 Plus
  family: opencodego-qwen3.5-plus-session-2026-05-21-b
  vendor: opencodego
  version: 1.2.0-session-2026-05-21-b-262144-52365
  maxInputTokens: 253952
  advertisedMaxOutputTokens: 8192
  advertisedContextWindow: 262144
  apiMaxOutputTokens: 52365
  metadataSource: models.dev
  supportsVision: true
  status: deprecated
  thinkingFamily: qwen
  configurationSchema: null
- mimo-v2-pro
  rawModelId: mimo-v2-pro
  name: OpenCode Go / Mimo V2 Pro
  family: opencodego-mimo-v2-pro-session-2026-05-21-b
  vendor: opencodego
  version: 1.2.0-session-2026-05-21-b-1048576-128000
  maxInputTokens: 1040384
  advertisedMaxOutputTokens: 8192
  advertisedContextWindow: 1048576
  apiMaxOutputTokens: 128000
  metadataSource: models.dev
  supportsVision: false
  status: deprecated
  thinkingFamily: mimo
  configurationSchema: null
- mimo-v2-omni
  rawModelId: mimo-v2-omni
  name: OpenCode Go / Mimo V2 Omni
  family: opencodego-mimo-v2-omni-session-2026-05-21-b
  vendor: opencodego
  version: 1.2.0-session-2026-05-21-b-262144-52365
  maxInputTokens: 253952
  advertisedMaxOutputTokens: 8192
  advertisedContextWindow: 262144
  apiMaxOutputTokens: 52365
  metadataSource: models.dev
  supportsVision: true
  status: deprecated
  thinkingFamily: mimo
  configurationSchema: null
- mimo-v2.5-pro
  rawModelId: mimo-v2.5-pro
  name: OpenCode Go / Mimo V2.5 Pro
  family: opencodego-mimo-v2.5-pro-session-2026-05-21-b
  vendor: opencodego
  version: 1.2.0-session-2026-05-21-b-1048576-128000
  maxInputTokens: 1040384
  advertisedMaxOutputTokens: 8192
  advertisedContextWindow: 1048576
  apiMaxOutputTokens: 128000
  metadataSource: models.dev
  supportsVision: false
  status: active
  thinkingFamily: mimo
  configurationSchema: null
- mimo-v2.5
  rawModelId: mimo-v2.5
  name: OpenCode Go / Mimo V2.5
  family: opencodego-mimo-v2.5-session-2026-05-21-b
  vendor: opencodego
  version: 1.2.0-session-2026-05-21-b-1000000-128000
  maxInputTokens: 991808
  advertisedMaxOutputTokens: 8192
  advertisedContextWindow: 1000000
  apiMaxOutputTokens: 128000
  metadataSource: models.dev
  supportsVision: true
  status: active
  thinkingFamily: mimo
  configurationSchema: null
- hy3
  rawModelId: hy3
  name: OpenCode Go / Hy3
  family: opencodego-hy3-session-2026-05-21-b
  vendor: opencodego
  version: 1.2.0-session-2026-05-21-b-256000-51136
  maxInputTokens: 247808
  advertisedMaxOutputTokens: 8192
  advertisedContextWindow: 256000
  apiMaxOutputTokens: 51136
  metadataSource: models.dev
  supportsVision: false
  status: active
  thinkingFamily: none
  configurationSchema: null
  limits: using bundled fallback
- hy3-preview
  rawModelId: hy3-preview
  name: OpenCode Go / Hy3 Preview
  family: opencodego-hy3-preview-session-2026-05-21-b
  vendor: opencodego
  version: 1.2.0-session-2026-05-21-b-256000-51136
  maxInputTokens: 247808
  advertisedMaxOutputTokens: 8192
  advertisedContextWindow: 256000
  apiMaxOutputTokens: 51136
  metadataSource: fallback
  supportsVision: false
  status: active
  thinkingFamily: none
  configurationSchema: null
- grok-4.5
  rawModelId: grok-4.5
  name: OpenCode Go / Grok 4.5
  family: opencodego-grok-4.5-session-2026-05-21-b
  vendor: opencodego
  version: 1.2.0-session-2026-05-21-b-500000-99936
  maxInputTokens: 491808
  advertisedMaxOutputTokens: 8192
  advertisedContextWindow: 500000
  apiMaxOutputTokens: 99936
  metadataSource: models.dev
  supportsVision: true
  status: active
  thinkingFamily: none
  configurationSchema: null
  limits: using bundled fallback

[ltmoerdani]

Hey @gwynnbleiidd, thanks for the report — the diagnostics you attached made this one straightforward to track down.

I spent some time tracing through the math, and here's what I found.

The problem is in estimateTokenCount(), a helper that guesses how many tokens a prompt will use before it hits the API. The heuristic it used — words * 1.15 — sounds reasonable for plain text, but JSON-serialized messages are anything but plain. Every {, }, ", :, , got counted as a separate "word". For structured conversations like yours, that meant the estimate was 3-5x higher than what the server actually saw.

How that plays out in practice:

Your conversation — roughly 754K tokens by the server's count — was being estimated at ~3.5M tokens locally. That's 3.5x the 1M context window. The safe-output calculation collapsed to max_tokens: 1, which is exactly what the diagnostics show: completion=1, finishReason=length. The model generated one token and stopped.

(Between you and me, I'm surprised this didn't surface sooner. The heuristic has been like this since the early days. I think most people just clear their chat when it starts feeling sluggish, not realizing the real trigger is a silent max_tokens: 1 behind the scenes.)

Fix is in PR #84. I replaced the word-count approach with something simpler: OpenAI's rule of thumb (1 token ≈ 4 characters), plus a 10% buffer for code-heavy prompts. It naturally overestimates for JSON, which is the safe direction. Also added a MIN_OUTPUT_BUDGET = 4096 floor so even if the estimate is off, the model can still generate a meaningful response.

Verified with a simulation matching your exact scenario:

• Before fix: max_tokens: 1 ❌
• After fix: max_tokens: 4,096 ✅
• Total budget used: ~76% of context window (safe)

Already shipped as v0.4.3 and installed locally. Would be great if you could pull the latest and see how it behaves on your end — especially curious if it holds up in longer sessions.