[BOT ISSUE] Together: stale pricing for deepseek-ai/DeepSeek-V4-Pro and missing pricing for moonshotai/Kimi-K2.6

[github-actions]

Summary

Two Together AI models in packages/proxy/schema/model_list.json have pricing gaps:

1. deepseek-ai/DeepSeek-V4-Pro — pricing is stale. The catalog lists $2.10/$4.40 (the original Together price when the model was added in May 2026 via issue [BOT ISSUE] Missing Together AI models: deepseek-ai/DeepSeek-V4-Pro, Qwen/Qwen3.5-397B-A17B #515), but Together AI has since reduced pricing to $1.74/$3.48.
2. moonshotai/Kimi-K2.6 — pricing is entirely absent. The entry has no input_cost_per_mil_tokens or output_cost_per_mil_tokens fields. Issue [BOT ISSUE] fix: add Together as provider for zai-org/GLM-5, moonshotai/Kimi-K2.5, moonshotai/Kimi-K2.6 #579 added together to its provider list but noted "the downstream fix job may want to add pricing" — it was never added.

Affected Entries

deepseek-ai/DeepSeek-V4-Pro (line ~3687)

Field	Catalog Value	Together Official
input_cost_per_mil_tokens	2.10	1.74
output_cost_per_mil_tokens	4.40	3.48
input_cache_read_cost_per_mil_tokens	0.20	0.20 (matches)
max_input_tokens	512000	512K (matches)

Current catalog entry:

"deepseek-ai/DeepSeek-V4-Pro": {
    "format": "openai",
    "flavor": "chat",
    "input_cost_per_mil_tokens": 2.1,
    "output_cost_per_mil_tokens": 4.4,
    "input_cache_read_cost_per_mil_tokens": 0.2,
    "displayName": "DeepSeek V4 Pro",
    "max_input_tokens": 512000,
    "available_providers": ["together"]
}

Note: The Fireworks entry (accounts/fireworks/models/deepseek-v4-pro) already has the correct $1.74/$3.48 pricing. The Together entry is the only stale one.

moonshotai/Kimi-K2.6 (line ~5595)

Field	Catalog Value	Together Official
input_cost_per_mil_tokens	(missing)	1.20
output_cost_per_mil_tokens	(missing)	4.50
input_cache_read_cost_per_mil_tokens	(missing)	0.20
multimodal	(missing)	true (text, image, video input)
max_input_tokens	(missing)	256000

Current catalog entry:

"moonshotai/Kimi-K2.6": {
    "format": "openai",
    "flavor": "chat",
    "displayName": "Kimi K2.6",
    "available_providers": ["baseten", "together"]
}

Verification Checklist

Check	Status	Detail
Cross-source (DeepSeek-V4-Pro)	✅	Pricing confirmed on Together serverless models ($1.74/$3.48) AND Together model page ($1.74/$3.48, cached $0.20)
Cross-source (Kimi-K2.6)	✅	Pricing confirmed on Together serverless models ($1.20/$4.50) AND Together model page ($1.20/$4.50, cached $0.20, 256K context, multimodal)
Already fixed	✅	Confirmed in model_list.json: DeepSeek-V4-Pro still at $2.10/$4.40; Kimi-K2.6 still has no pricing
Duplicate check	✅	No open issue covers these pricing gaps. Prior issue #515 (closed) originally added DeepSeek-V4-Pro at $2.10/$4.40. Prior issue #579 (closed) added Together provider for Kimi-K2.6 but noted pricing was absent.

Verification Notes

Field	Source	Notes
DeepSeek-V4-Pro $1.74 input	Together serverless models page + model detail page	Both pages agree
DeepSeek-V4-Pro $3.48 output	Together serverless models page + model detail page	Both pages agree
DeepSeek-V4-Pro $0.20 cached input	Together model detail page	Already correct in catalog
Kimi-K2.6 $1.20 input	Together serverless models page + model detail page	Both pages agree
Kimi-K2.6 $4.50 output	Together serverless models page + model detail page	Both pages agree
Kimi-K2.6 $0.20 cached input	Together model detail page	Not in catalog
Kimi-K2.6 multimodal	Together model detail page	"supports text, image, and video input" with MoonViT vision encoder
Kimi-K2.6 256K context	Together model detail page	"256K token context window"
Kimi-K2.6 max_output_tokens	Not published	Neither Together page specifies max output tokens — omitted

Local Files Inspected

• packages/proxy/schema/model_list.json — confirmed both entries and their current values
• packages/proxy/scripts/sync_models.ts — confirmed neither model is in SYNC_PRESERVED_FIELDS

{
  "kind": "cost_update",
  "provider": "together",
  "models": ["deepseek-ai/DeepSeek-V4-Pro", "moonshotai/Kimi-K2.6"],
  "status": "active",
  "model_specs": {
    "deepseek-ai/DeepSeek-V4-Pro": {
      "format": "openai",
      "flavor": "chat",
      "input_cost_per_mil_tokens": 1.74,
      "output_cost_per_mil_tokens": 3.48,
      "input_cache_read_cost_per_mil_tokens": 0.2,
      "displayName": "DeepSeek V4 Pro",
      "max_input_tokens": 512000,
      "available_providers": ["together"]
    },
    "moonshotai/Kimi-K2.6": {
      "format": "openai",
      "flavor": "chat",
      "multimodal": true,
      "input_cost_per_mil_tokens": 1.2,
      "output_cost_per_mil_tokens": 4.5,
      "input_cache_read_cost_per_mil_tokens": 0.2,
      "displayName": "Kimi K2.6",
      "max_input_tokens": 256000,
      "available_providers": ["baseten", "together"]
    }
  },
  "source_urls": [
    "https://docs.together.ai/docs/serverless-models",
    "https://www.together.ai/models/deepseek-v4-pro",
    "https://www.together.ai/models/kimi-k26"
  ]
}

[github-actions]

Prepared catalog updates for deepseek-ai/DeepSeek-V4-Pro, moonshotai/Kimi-K2.6. Updated models: deepseek-ai/DeepSeek-V4-Pro, moonshotai/Kimi-K2.6.

Verification sources: https://docs.together.ai/docs/serverless-models, https://www.together.ai/models/deepseek-v4-pro, https://www.together.ai/models/kimi-k26
Verified metadata applied:

• deepseek-ai/DeepSeek-V4-Pro: format=openai, flavor=chat, displayName=DeepSeek V4 Pro, max_input_tokens=512000, max_output_tokens=not provided, pricing(input/output per 1M)=1.74/3.48, input_cache_read_cost_per_mil_tokens=0.2, providers=together
• moonshotai/Kimi-K2.6: format=openai, flavor=chat, displayName=Kimi K2.6, max_input_tokens=262144, max_output_tokens=not provided, pricing(input/output per 1M)=1.2/4.5, input_cache_read_cost_per_mil_tokens=0.2, providers=baseten|together, multimodal=true

Included in the daily batch PR: #822