Go
Haider Go
Hosted open-weight models behind one subscription - the shortcut, not a requirement.
Models#
Fifteen hosted open-weight models - one subscription, one endpoint.
| Model | Lane | Input | Cached | Output | Context | Speed advertised | Cache | Availability |
|---|---|---|---|---|---|---|---|---|
glm-4.7-flashx | Economy | $0.21 | $0.03 | $1.20 | unknown | - | cached input supported | live |
qwen3.5-flash | Economy | $0.30 | $0.06 | $1.20 | unknown | - | cached input supported | live |
glm-5.3-flash | Fast coding | $0.45 | $0.09 | $1.50 | 200k | - | cached input supported | live |
deepseek-v4-flash | Default | $0.66 | $0.02 | $1.98 | 128k | - | cached input supported | live |
minimax-m3 | Long context | $0.90 | $0.18 | $3.60 | 1M | - | cached input supported | live |
qwen3-coder-next | Coding | $0.90 | $0.18 | $4.50 | unknown | - | cached input supported | live |
qwen3.7-plus | Balanced | $1.20 | $0.24 | $4.80 | 262k | - | cached input supported | live |
qwen3.8-27b | Compact | $1.20 | $0.12 | $9.00 | 1M | - | cached input supported | live |
kimi-k2.7-code | Premium coding | $2.85 | $0.57 | $12.00 | 262k | - | cached input supported | live |
deepseek-v4-pro | Default+ | $3.96 | $0.13 | $11.88 | 128k | - | cached input supported | live |
glm-5.2 | Premium planner | $4.20 | $0.78 | $13.20 | 200k | - | cached input supported | live |
glm-5.3 | Premium coding | $4.20 | $0.78 | $13.20 | 200k | - | cached input supported | live |
qwen3.8-2.4t-a95b | Open flagship | $6.00 | $0.60 | $18.00 | 1M | - | cached input supported | live |
qwen3.8-max | Frontier | $6.00 | $0.75 | $18.00 | 1M | - | cached input supported | live |
kimi-k3 | Flagship | $9.00 | $0.90 | $45.00 | 1M | - | cached input supported | live |
Rates per 1M tokens. Cached input meters at the cached rate automatically; usage is drawn from your weekly allowance first, then usage-credits. Context is the window the model publishes; speed is the upstream's advertised throughput, not our measurement.
GET /v1/models lists what is live on this deployment right now; models roll out as upstream lanes are switched on.
GET /v1/account returns your own plan and remaining weekly allowance, authenticated with the same key as inference - so a harness can show your status without a browser. It refreshes once a minute; calls inside that window return the cached snapshot with "cached": true.
curl https://haidercode.ai/v1/account \ -H "Authorization: Bearer $HAIDER_GO_KEY"
Plans#
| Plan | Price | Allowance |
|---|---|---|
| Go | $40/month | Starter - a limited weekly allowance |
| Go Max | $200/month | 20× the usage of Go |
Allowances are metered over a rolling 7-day window, not a calendar week - there is no reset day to game.
Both plans include artifacts: publish HTML pages from a session and share one link. See Artifacts.
Both plans include one weekly allowance for hosted models and tools; usage-credits top it up. See Hosted tools for discovery, invocation and funding.
usage-credits#
usage-credits are dollars, not mystery units. When the weekly allowance runs out, spend continues from your usage-credits balance. Top up any time - see Billing & usage-credits.
The harness stays free#
Haider itself is free with your own provider keys. Go is the shortcut for people who would rather not manage keys - never a requirement.