PlanetGPU
Open Model Watch · verified 2026-06-18

Frontier open and low-cost AI models for builders

Track models where the buyer decision actually matters: hosted API price, coding-agent support, self-host confidence, GPU alternative cost, source freshness, and affiliate disclosure.

Browse VRAM catalogCompare coding agentsOfficial sources preferredNo review/rating schemaAffiliate links disclosed
LLM model requirements represented by transformer blocks, parameter volumes, and GPU memory stacks.
Models tracked
7
Open weight
1
confirmed rows
High pricing
5
transparent API rows
Affiliate rows
4
disclosed CTAs

Open Model Watch is buyer intelligence, not generic rankings

PlanetGPU does not assign model quality ratings here. Rows compare source-backed buyer facts: price transparency, context window, coding-tool support, open/self-host confidence, GPU-fit notes, and caveats that affect real purchasing decisions.
Featured

GLM-5.2

A strong open-model candidate for coding-agent workflows where the buyer choice is coding subscription, hosted API, or H200/B200-class self-hosting.

medium pricingavailable
Featured

MiniMax-M3

A coding and multimodal model route with Token Plan tiers, Claude Code/Cursor compatibility, and separately published pay-as-you-go API prices.

high pricinguncertain
Featured

MiMo V2.5 Pro

A hosted Xiaomi MiMo buyer page for coding-tool Token Plan tiers, pay-as-you-go API cost, and invite-code caveats.

high pricingnot-available
Featured

Qwen3.7 Max / Qwen AI Coding Plan

A Qwen3.7 opportunity row for builders comparing Alibaba Cloud's AI Coding Plan, Qwen Code support, Model Studio credits, and self-host alternatives in the Qwen open-model family.

medium pricinguncertain

Open Model Watch

7 of 7 model opportunities shown. Filter by model status, self-host confidence, API pricing transparency, and affiliate availability.

GLM-5.2
Z.ai · verified 2026-06-18
open-weightavailablemedium pricing

A strong open-model candidate for coding-agent workflows where the buyer choice is coding subscription, hosted API, or H200/B200-class self-hosting.

API pricing

Usage based; verify current GLM-5.2 token prices on Z.ai or provider docs before volume commitments.

Coding / subscription

GLM Coding Plan from $18/month via Z.ai. Claude Code, Cline, Kilo Code, OpenCode, OpenClaw/Clawdbot, Cursor, Windsurf, Trae, and other agent tools advertised by Z.ai.

GPU fit

Open-weight deployment is the main PlanetGPU angle; treat long-context production serving as H200/B200-class multi-GPU planning, not a small local run.

open-weightcoding-agentself-hostaffiliate disclosed
View buyer page
MiniMax-M3
MiniMax · verified 2026-06-18
hybriduncertainhigh pricing

A coding and multimodal model route with Token Plan tiers, Claude Code/Cursor compatibility, and separately published pay-as-you-go API prices.

API pricing

Standard <=512K: $0.30/M input, $1.20/M output, $0.06/M cache read. Higher rates apply above 512K and priority tier.

Coding / subscription

Token Plan Plus $20/month, Max $50/month, Ultra $120/month. Claude Code, Cursor, OpenClaw, TRAE, Hermes Agent, MiniMax CLI, MiniMax Code, Anthropic-compatible API, and OpenAI-compatible API.

GPU fit

Treat self-hosting as a verification step until official weights, license terms, and serving requirements are confirmed.

coding-agentmultimodalsubscriptionaffiliate disclosed
View buyer page
MiMo V2.5 Pro
Xiaomi MiMo · verified 2026-06-18
hosted-apinot-availablehigh pricing

A hosted Xiaomi MiMo buyer page for coding-tool Token Plan tiers, pay-as-you-go API cost, and invite-code caveats.

API pricing

mimo-v2.5-pro overseas API: $0.0036/M cache-hit input, $0.435/M cache-miss input, $0.87/M output.

Coding / subscription

Token Plan tiers shown by PlanetGPU: Lite $6/month, Standard $16/month, Pro $50/month, Max $100/month. OpenClaw, OpenCode, MiMo Claw, MiMo Code, and supported programming-tool integrations.

GPU fit

Compare GPU rental only against alternative self-hostable models; do not assume MiMo V2.5 Pro weights are downloadable.

hosted-apicoding-agentaffiliateaffiliate disclosed
View buyer page
Kimi K2.7 Code
Moonshot AI / Kimi · verified 2026-06-18
hosted-apiuncertainhigh pricing

A coding-specialized hosted API model with official Markdown pricing, 262K context, tool calling, JSON mode, and a high-speed variant.

API pricing

kimi-k2.7-code: $0.19/M cache-hit input, $0.95/M cache-miss input, $4.00/M output. Highspeed doubles those rates.

Coding / subscription

No PlanetGPU-tracked subscription plan; API billing is the primary buyer path. Coding model with tool calls, JSON mode, partial mode, context caching, and Kimi API compatibility.

GPU fit

Treat as hosted API by default; compare GPU rental only if official open weights and serving requirements are available.

coding-agenthosted-apipricing-transparent
View buyer page
Qwen3.5 Plus / Flash
Alibaba Cloud Model Studio · verified 2026-06-18
hosted-apiavailablehigh pricing

A Qwen API buyer entry for long-context Model Studio pricing, useful when comparing low-cost hosted inference against self-hosted alternatives.

API pricing

Global deployment examples: Qwen3.5-Plus min input $0.115/M, output $0.688/M; Qwen3.5-Flash min input $0.029/M, output $0.287/M.

Coding / subscription

No PlanetGPU-tracked coding subscription; Model Studio API pricing is the primary buyer path. OpenAI-compatible Model Studio API can be used through gateways; direct coding-agent plan support is not the main Qwen buyer signal here.

GPU fit

Use GPU comparisons when choosing open Qwen-family self-hosted alternatives; API deployment-mode prices vary by region.

Sources
Official
hosted-apiopen-familylong-context
View buyer page
Qwen3.7 Max / Qwen AI Coding Plan
Alibaba Cloud / Qwen · verified 2026-06-18
hybriduncertainmedium pricing

A Qwen3.7 opportunity row for builders comparing Alibaba Cloud's AI Coding Plan, Qwen Code support, Model Studio credits, and self-host alternatives in the Qwen open-model family.

API pricing

Alibaba benefits page advertises 70,000,000+ free Qwen AI tokens for Model Studio trials and an AI Token Plan at $30/seat with fixed credits. Per-token Qwen3.7-Max pricing should be verified in-console before spend.

Coding / subscription

Alibaba AI Coding Plan campaign lists a maximum of 90,000 requests/month, support for qwen3-coder-plus, and fixed-price coding-plan positioning; static page did not expose a simple monthly dollar price. AI Coding Plan page lists Qwen Code, OpenClaw, OpenCode, Claude Code, Claude Code IDE plugin, Codex, Cline, Cursor, Kilo CLI, and Kilo Code IDE plugin.

GPU fit

Treat Qwen3.7-Max as hosted/API or coding-plan first. For self-host comparisons, verify the exact Qwen3.7 open-weight release, license, parameter count, and serving requirements before mapping to H100/H200/B200 GPU rental.

coding-agentqwenaffiliatemodel-studioaffiliate disclosed
View buyer page
DeepSeek V4 Pro / Flash
DeepSeek · verified 2026-06-18
hosted-apiuncertainhigh pricing

A high-transparency hosted API row for DeepSeek V4 Flash and Pro, with OpenAI and Anthropic-compatible base URLs and 1M context in official docs.

API pricing

deepseek-v4-pro: $0.003625/M cache-hit input, $0.435/M cache-miss input, $0.87/M output. deepseek-v4-flash: $0.0028/M cache-hit input, $0.14/M cache-miss input, $0.28/M output.

Coding / subscription

No PlanetGPU-tracked subscription; API balance billing is the buyer route. OpenAI-format and Anthropic-format base URLs; supports thinking mode, JSON output, tool calls, and FIM in non-thinking mode.

GPU fit

Use hosted API economics as the default; self-host planning needs official weight/license and serving requirement confirmation.

Sources
Official
hosted-apipricing-transparentlong-context
View buyer page

Google Search Console checklist

After deploy, inspect and request indexing for /open-models, /model/glm-5-2, /model/minimax-m3, and /model/mimo-v2.5-pro, and /model/qwen3-7-max. Re-submit sitemap.xml and confirm structured data parses without review/rating schema.