GLM vs DeepSeek: API Cost Compared (2026)

At the cheapest usable tier, DeepSeek V4-Flash ($0.14/$0.28 official) undercuts GLM, but GLM-4.5-Air (measured at $0.13/$0.85 via OpenRouter) wins on input price, so the real answer depends on whether your workload is input-heavy or output-heavy. This comparison keeps the prices honest about where each number comes from, because the two models reach us through different doors.

The asymmetry matters before we quote a single figure. DeepSeek we call on its own official endpoint, so its rates are first-party. GLM we reach through OpenRouter, not a native Z.ai key, so every GLM price below is the OpenRouter-routed value and every official Z.ai figure is sourced research pending native re-verification. Mixing the two without saying which is which is the single most common mistake in GLM-vs-DeepSeek posts, and it quietly inflates or deflates the comparison by 20 to 40 percent.

GLM vs DeepSeek API cost (verified 2026-06)

Here is the rate card both ways, with the source of each number labeled. All figures are USD per million tokens.

Model	Input ($/1M)	Output ($/1M)	Source
DeepSeek V4-Flash	$0.14	$0.28	Official api.deepseek.com
GLM-4.6	$0.43	$1.74	Measured via OpenRouter
GLM-4.6 (official)	$0.60	$2.20	According to Z.ai pricing
GLM-4.5-Air	$0.13	$0.85	Measured via OpenRouter
GLM-4.5-Air (official)	$0.20	$1.10	According to Z.ai pricing

The pattern jumps out immediately. On both GLM tiers, the OpenRouter-routed price is cheaper than the official Z.ai price, the same OpenRouter-undercuts-official behavior the cluster documented for Qwen. So when you compare GLM against DeepSeek, you have to decide which GLM price you are actually going to pay. If you route through OpenRouter, GLM-4.5-Air at $0.13 input edges out DeepSeek's $0.14. If you go native on Z.ai, GLM-4.5-Air costs $0.20 input and DeepSeek is clearly cheaper.

Output is where DeepSeek pulls ahead regardless of routing. According to DeepSeek API Docs, V4-Flash bills output at $0.28 per million. GLM-4.5-Air's $0.85 output (OpenRouter) is roughly 3x that, and GLM-4.6's $1.74 is over 6x. For chatbots and agents that generate long responses, output dominates the bill, and that tilts the math decisively toward DeepSeek.

Where GLM-4.5-Air actually competes

GLM-4.5-Air is a lightweight GLM tier that trades some capability for a near-DeepSeek input price. The reason to consider it is narrow but real: workloads that are input-heavy and output-light. Think classification, extraction, reranking, or RAG pipelines that feed in large retrieved context and emit a short verdict.

In that shape, the input rate is most of your bill, and GLM-4.5-Air's $0.13 input (measured via OpenRouter) is the lowest number in the table. The catch is the output multiplier. The moment your output grows, DeepSeek's $0.28 output reclaims the lead, because $0.85 per million output adds up fast on anything conversational.

According to Z.ai pricing, GLM-4.5-Air's official rate is $0.20/$1.10, which is the figure to budget against if you hold a native Z.ai key rather than routing through a third party. We mark this needs_native_reverify because we have not yet re-confirmed it on a first-party key, and the OpenRouter-versus-official gap is exactly the kind of thing that drifts.

One thing DeepSeek has no answer for: a free Flash tier

There is one GLM advantage that no DeepSeek price can match. According to Z.ai pricing, GLM-4.7-Flash and GLM-4.5-Flash run on a genuine free tier on the Z.ai international platform, zero cost in and out. DeepSeek has no free tier; its cheapest real rate is still $0.14 input.

This is an official-Z.ai fact, not an OpenRouter one. OpenRouter actually charges for the same Flash model (its catalog lists GLM-4.7-Flash at $0.06/$0.40), so the free tier only exists if you go native on Z.ai. For prototyping, low-volume side projects, or evaluation before you commit, the free Flash tier is a legitimate reason to start on GLM rather than DeepSeek, even though DeepSeek wins on paid throughput.

First-hand: we ran the same coding prompt on both

Documentation gives you rates. We wanted to see the two models answer the same kind of task, so we ran a coding prompt against each and recorded tokens, latency, and billed cost. The GLM call went through OpenRouter; the DeepSeek call hit the official endpoint. That routing asymmetry is real and we are not hiding it: it is exactly why the prices above carry different source labels.

Model	Prompt tokens	Output tokens	Latency	Routing
GLM-4.6 (coding)	30	165	3.78s	OpenRouter
DeepSeek V4-Flash (coding)	27	120	2.25s	Official api.deepseek.com

The GLM-4.6 coding call billed $0.0002972 for 30 input and 165 output tokens, measured via OpenRouter. DeepSeek answered the same class of prompt in 2.25 seconds against GLM's 3.78, so DeepSeek was both faster and, at $0.14/$0.28, cheaper per token on this run. GLM-4.6 produced a longer answer (165 tokens versus 120), which is partly why it cost more, but even normalized per token the OpenRouter GLM-4.6 rate sits well above DeepSeek's. The honest read: on this single coding sample, DeepSeek was the better cost-per-result, while GLM-4.6's longer output may suit tasks where more detail is the point.

We treat one run as a signal, not a verdict, and we disclose that the GLM number is OpenRouter-routed rather than native Z.ai. A native-key re-run could shift GLM's effective price toward the higher official rate, which would widen, not narrow, DeepSeek's edge.

The verdict in one line per use case

Choose DeepSeek V4-Flash when output volume is significant or you want a single first-party endpoint and the cheapest output rate at $0.28. Choose GLM-4.5-Air via OpenRouter when your workload is input-heavy and output-light, where its $0.13 input is the floor. Start on GLM Flash's free tier when you are prototyping and have not committed budget yet. For everything output-heavy, DeepSeek's pricing structure is hard to beat.

For the full GLM rate card across every tier, see the GLM API pricing hub. For the DeepSeek side in depth, see the DeepSeek API pricing hub.

FAQ

Is GLM cheaper than DeepSeek? It depends on routing and workload. On input, GLM-4.5-Air at $0.13/M (measured via OpenRouter) just undercuts DeepSeek V4-Flash's $0.14/M. On output, DeepSeek's $0.28/M beats every GLM tier, so for output-heavy work DeepSeek is cheaper overall. On official Z.ai pricing, GLM-4.5-Air costs $0.20/M input, and DeepSeek is cheaper across the board.

Why are GLM's OpenRouter prices different from Z.ai's official prices? GLM is reached here through OpenRouter, not a native Z.ai key. OpenRouter's routed prices ($0.43/$1.74 for GLM-4.6, $0.13/$0.85 for GLM-4.5-Air) come in below the official Z.ai rates ($0.60/$2.20 and $0.20/$1.10). Always confirm which path you are billing through before budgeting.

Does GLM have a free tier that DeepSeek doesn't? Yes. According to Z.ai pricing, GLM-4.7-Flash and GLM-4.5-Flash are free on the Z.ai international platform, zero cost in and out. DeepSeek has no free tier; its cheapest rate is $0.14/M input. Note the free tier is a native Z.ai feature, not an OpenRouter one.

Which is faster, GLM-4.6 or DeepSeek V4-Flash? On our coding test, DeepSeek V4-Flash answered in 2.25 seconds against GLM-4.6's 3.78 seconds, with the GLM call routed via OpenRouter. DeepSeek was faster on that run, though GLM-4.6 returned a longer answer.

This is part of the GLM API pricing hub.

Author: Kevin Fan, Customer Success Manager at China LLM Directory, specializing in Chinese LLM ecosystem pricing. Last verified: 2026-06-26.