GLM-5.3 Flash wins on price ($0.15 vs $0.2/1M input). GPT-5.6 Luna wins on coding (88 vs 82) and writing quality. For most workflows, GPT-5.6 Luna is the stronger default — best budget model from a frontier lab — near-frontier scores at commodity price.
The shortest way to see the safest default, the lower-cost option, and the specialist pick before you read deeper.
GPT-5.6 Luna is the safest overall answer here when you want the strongest default instead of the lowest list price.
Mistral: Mistral Nemo is the lower-cost option to start with when you still need useful output at scale.
GLM-5.3 Flash is the better pick when response speed matters more than maximum reasoning depth.
GPT-5.6 Luna leads on coding with a score of 88 vs 82 for GLM-5.3 Flash.
GPT-5.6 Luna has the larger context window: 1.05M vs 1M for GLM-5.3 Flash.
GLM-5.3 Flash is cheaper at $0.15/1M input tokens vs $0.2/1M for GPT-5.6 Luna.
Go with GPT-5.6 Luna if you want one model to handle coding and writing — it targets cheap high-throughput summarization, drafting, and routine agent steps.
GLM-5.3 Flash earns its place when your work is mostly cheap multimodal work at scale on MIT-licensed weights, even though it loses the overall count here.
GLM-5.3 Flash is the more cost-efficient option at $0.15/1M input — GPT-5.6 Luna costs 1x more per input token, so the gap is worth taking seriously wherever token volume rather than peak quality drives the bill.
Switch the scoring lens to see whether the top answer changes when you care more about cost, speed, or long-document work.
Z.ai / Budget / Aug 27, 2026
Native vision and video, MIT weights, fifteen cents per million.
Ranks models by the broadest mix of coding, writing, research, and long-context usefulness.
You need top-tier reasoning or a published SWE-bench Verified figure — this is a volume model, not a ceiling model.
The fastest way to see where the recommendation shifts when your priority changes.
Native vision and video, MIT weights, fifteen cents per million.
Best budget model from a frontier lab — near-frontier scores at commodity price.
Every figure below is the provider's list price or a published capability score — the same numbers the recommendation on this page is built from.
| Model | Input | Output | Est. month | Context | Speed | Coding | Writing | Research |
|---|---|---|---|---|---|---|---|---|
| GPT-5.6 LunaOpenAI | $0.20/1M | $1.20/1M | $4.40 | 1.1M tokens | Fast | 88 | 85 | 84 |
| GLM-5.3 FlashZ.ai | $0.15/1M | $0.50/1M | $2.50 | 1M tokens | Fast | 82 | 76 | 78 |
Capability scores are out of 100 and reflect our own weighting of published benchmarks and production signals — see how we evaluate models. “Est. month” assumes 10M input and 2M output tokens at list price, with no batch or caching discounts applied, so treat it as a ceiling.
What each one is genuinely good at, where it falls down, and the situations we would steer you away from it — not just the headline score.
The small, fast, cheap tier of the GPT-5.6 family — near-frontier scores on many benchmarks at commodity pricing after its ~80% July price cut.
Your workload actually uses the long context window — recall drops to 41% past 512K tokens.
The budget disruptor of 2026. After the price cut, Luna delivers benchmark scores that embarrass models 10x its price. Just don't trust it with genuinely long context — recall collapses past 512K tokens.
Fully public July 9, 2026; price cut ~80% to $0.20/$1.20 on July 30, 2026 (launched at $1/$6). Many third-party pages still show the old price.
A 320B-A18B mixture-of-experts model with native vision and video, MIT-licensed weights and a 1M context — at fifteen cents per million input tokens.
You need top-tier reasoning or a published SWE-bench Verified figure — this is a volume model, not a ceiling model.
The cheapest genuinely multimodal 1M-context model worth using in August 2026. If you are running vision or video at volume and can host weights, nothing at this price is close.
Released August 26, 2026. 320B total parameters, 18B active (320B-A18B MoE). Z.ai lists $0.15/1M input, $0.03/1M cached input and $0.50/1M output, with a launch promotion halving those rates through September 9, 2026. Self-reported against GLM-5.2: DeepSWE 63.4 vs 46.2, AutomationBench 48.8 vs 26.2.
UseRightAI recommendations are based on practical decision factors people actually feel in day-to-day use.
Newsletter
Useful if you care about ranking shifts, pricing changes, or a better recommendation appearing in this decision path.
No spam. Useful updates only. Affiliate disclosures always clearly labeled.
GPT-5.6 Luna wins on more of the categories we score — coding, writing, budget — so it is the better default of the two. GLM-5.3 Flash is the better pick when your work is mostly cheap multimodal work at scale on MIT-licensed weights. Neither is universally "better": GPT-5.6 Luna is aimed at cheap high-throughput summarization and drafting, GLM-5.3 Flash at cheap multimodal work at scale on MIT-licensed weights.
GLM-5.3 Flash is cheaper at $0.15/1M input and $0.5/1M output. GPT-5.6 Luna costs $0.2/1M input and $1.2/1M output.
GPT-5.6 Luna has the larger context window at 1.05M tokens vs GLM-5.3 Flash's 1M. For large document analysis, GPT-5.6 Luna is the stronger pick.
GPT-5.6 Luna is better for coding with a score of 88 vs GLM-5.3 Flash's 82 (out of 100). Claude Fable 5 is the overall coding leader in this directory at 100/100.
Both GLM-5.3 Flash and GPT-5.6 Luna have similar speed profiles — rated fast. Neither will be the bottleneck if latency is your deciding factor.
Long-context recall collapses at scale: 41.3% on 512K–1M token tasks vs Terra's 72.5%. Text and image input only — no video, audio, or native PDF ingestion like Gemini 3.6 Flash. Avoid it if your workload actually uses the long context window — recall drops to 41% past 512K tokens. That is the main case for looking at GLM-5.3 Flash instead.
No published SWE-bench Verified score. The launch promotion halves these rates only through September 9, 2026 — the $0.15/$0.50 list price applies after that. Avoid it if you need top-tier reasoning or a published SWE-bench Verified figure — this is a volume model, not a ceiling model. Against GPT-5.6 Luna specifically, the gap shows up most on coding (88 vs 82).
Take a moderate workload of 10M input and 2M output tokens a month. GLM-5.3 Flash runs $2.50 (at $0.15/1M in and $0.5/1M out); GPT-5.6 Luna runs $4.40 (at $0.2/1M in and $1.2/1M out). That is a $1.90/month difference — GLM-5.3 Flash is the cheaper of the two at this volume, and the gap scales linearly as you send more. Output tokens dominate the bill on both, so prompt length matters far less than response length.
Yes, and for most teams that beats picking one. A common split is GPT-5.6 Luna for cheap high-throughput summarization and drafting, with GLM-5.3 Flash handling cheap multimodal work at scale on MIT-licensed weights. Routing high-volume, low-stakes calls to GLM-5.3 Flash at $0.15/1M and reserving GPT-5.6 Luna for the hard cases is usually the cheapest arrangement that does not cost you quality.